A new Shopify setup page guides you through moving each connected store to DelightChat app embeds. Each store gets its own checklist, with steps to reconnect Shopify or enable the required embeds in your live theme. Complete the checks to keep your storefront features ready for Shopify’s updated requirements.
Stores that need to move their storefront features to Shopify app embeds now have a setup checklist inside DelightChat. The page shows the stores that still need attention and the steps for each one. You can see whether to reconnect Shopify or enable an app embed, instead of checking each integration and theme setting separately.
Follow the checklist for each store shown. If no stores need attention, the page shows that you are all set. Checking the live theme helps you complete the setup for the storefront features you use.
Enable retries when creating a WhatsApp broadcast, and three retry dates are filled in automatically, one day apart at the same time of day as your broadcast. Edit or remove any retry to suit your campaign instead of setting every date by hand.
Retry Broadcast gives another delivery attempt to recipients who did not receive a broadcast because of Meta's frequency cap. This update fills in the retry schedule when you enable the feature, so you can review a ready-made set of dates instead of entering each one yourself. It suggests three retries, one, two, and three days after the original send, at the same time of day.
Existing custom retry dates stay in place. Retries provide another delivery attempt; they do not guarantee delivery or resend the campaign to every recipient. For the full campaign setup, see How to send promotional broadcasts via WhatsApp.
Use Test a message in Inbox Workflows to enter a sample customer message and see which workflows would run. Results appear in the flowchart and list, so you can check the path before relying on it. No messages are sent to customers.
You can now check how a customer message would move through your inbox workflows without starting a customer conversation. Try a greeting, an order question, or a WhatsApp menu choice and follow the resulting path. This helps your team review which replies, assignments, and other actions a workflow would produce before using it with customers.
This is a simulation: it sends no customer messages and makes no changes to live tickets. It helps you check workflow logic, but it does not confirm message delivery or whether an agent is online.
Related setup for WhatsApp button workflows: Create a WhatsApp Chatbot.
The new Flow View in Inbox Workflows shows how your workflows run for each channel, with a zoomable flowchart and a list view. Follow conditions, actions and WhatsApp button branches, then open a workflow to review its details. A problems panel highlights dead-end buttons and deleted agents so you can spot issues before they affect customers.
If you use several workflows on the same channel, it can be hard to see how they fit together from individual settings screens. Flow View brings the paths into one place. For example, you can follow a WhatsApp menu button to the workflow that handles the customer's next reply, then review the conditions and actions along that path.
Flow View is read-only. Reviewing it does not change your workflows; use the workflow settings to make edits after you identify what needs attention.
Related setup for WhatsApp button workflows: Create a WhatsApp Chatbot.
You can now paste images and files, or drag them into the reply box, when chatting on WhatsApp, Instagram DMs, Facebook Messenger and Live Chat. Attach screenshots and documents directly from your clipboard or desktop without opening the file picker.
When a customer needs a product photo, a screenshot, or a document, your team can add it straight to the conversation from the clipboard or desktop. The attachment appears in the reply box, where you can check it before sending. This removes the file-picker step from replies on the four supported channels.
There is no separate setting to enable. The channel's existing file-type and size restrictions still apply. For help with replying and attachments, see the Support Inbox guide.
Build smarter segments for WhatsApp broadcasts. Start from a saved Shopify segment, filter by Shopify customer attributes, or re-target people who received, read, or did not get a past broadcast.
Then send your next campaign to exactly that group.
Choose an audience based on Shopify customer data or the response to an earlier broadcast. You can link a segment you already created in Shopify, build a new segment in DelightChat, or create a follow-up audience from a past campaign's delivery and read results. This gives you separate audiences for people who read an offer and people who did not receive it, instead of using the same list for every follow-up.
Open Customer Segments and select Create Segment. Choose Shopify Customers to sync an existing Shopify segment or build a new one; this source requires a connected Shopify store. For a campaign follow-up, choose the re-target source, select the past broadcast and target condition, then save the segment. Select that audience when creating your next broadcast. See how segment filters work for the general filtering and audience setup.
When a shopper reports a wrong, missing, or damaged order, the AI chatbot now escalates the chat to a human agent right away. Your team handles these complaints directly, instead of the bot trying to resolve them alone.
A report of a wrong item, an item missing from the order, or damaged goods needs your team to investigate and decide what to do next. AI Agent now treats these complaints as a reason to transfer the conversation, rather than continuing with an automated answer.
This also covers a replacement or reshipment request caused by one of those defects. A shopper who only asks about your replacement policy can still receive a policy answer when that topic is enabled and your knowledge base has the information.
Open AI Agent settings, select the support channel, and check Auto Transfer. Choose the destination for handed-off tickets and the customer-facing handoff message, then save. The complaint detection uses that existing handoff configuration.
Related guide: Work Alongside Your AI Agent.
If a shopper says the chatbot's earlier answer was wrong, the AI chatbot now hands the chat to a human agent instead of repeating itself. Shoppers who push back reach your team faster.
When a customer contradicts an answer AI Agent already gave, continuing to repeat that answer can leave the conversation stuck. The chatbot now treats that dispute as a reason for your team to take over and review the customer’s concern.
The distinction is between disputing a previous claim and making a new request. A new preference or a follow-up question does not automatically count as a dispute. Your human agent can read the conversation and correct the answer or investigate the issue.
Check the selected channel’s Auto Transfer settings in AI Agent settings. The transfer follows your existing ticket assignment rule and handoff-message configuration.
Related guide: Work Alongside Your AI Agent.
The template dropdown in reply, broadcast, and automations is now wider and shows each template's approval date. Long names no longer get cut off, and you can hover a truncated name to see it in full. The date helps you tell near-identical templates apart.
See more of a template's name before choosing it, and use its approval date to distinguish similar versions. This is useful when your team keeps several approved templates for the same offer, language or workflow. You can inspect a truncated name instead of guessing which version a short label refers to.
Open the template selector when composing a reply, creating a broadcast or configuring an automation. Review the approval date beside the template, and hover over a shortened name to see the full label before selecting it. The updated picker appears automatically in these flows. For the campaign steps around template selection and configuring its variables, see the broadcast setup guide.
The conversation export tool now accepts any date range, not just up to six months. It splits a large range into parts automatically and skips periods you already exported.
Admins can choose a longer period in one request instead of manually submitting each smaller range. DelightChat plans separate export files for the parts that need exporting and skips dates that were already requested for the channel.
Open Reporting → Conversation Export, choose the channel and your full date range, then review the number of exports shown before submitting. Track each resulting request in the history table and download the completed files before expiry. The screen still applies safeguards to how much one submission can create.
The chatbot now correctly recognises when a shopper accepts a branch-check or store-location offer. It stops re-asking the same question and no longer matches against deleted products.
When AI Agent offers to check a branch and the shopper replies “yes”, that reply now continues the availability flow. The chatbot can ask which branch is wanted or use a location the shopper supplies, instead of repeatedly making the same offer.
The product retrieval also checks the available store data rather than relying on a product that has been deleted. This keeps the follow-up connected to the product the shopper is discussing and the location information the bot can actually use.
For location-aware availability, keep Share location-wise availability on in the channel’s AI Agent settings. Keep Shopify products and locations current. The follow-up handling does not add another setting.
The chatbot sometimes stopped replying right after a shopper confirmed a product or order selection. It now keeps the conversation going as expected.
A short confirmation can still carry an important choice. When a shopper selects a product, variant, or order that the conversation has just discussed, AI Agent now treats the selection as part of the active request instead of dismissing it as a simple acknowledgement.
The chatbot can continue with the relevant product or order information when it has evidence to answer. An ordinary “thanks” remains different from a customer committing to a specific option, so the improvement does not turn every acknowledgement into another detailed reply.
This applies within the existing AI Agent conversation flow. Keep the channel’s enabled capabilities and data sources up to date in AI Agent settings; no separate confirmation setting is required.
Related guide: What Your AI Agent Can and Can’t Help With.
Products with very long or HTML-heavy descriptions never got indexed for AI, so the chatbot could not see them. They are now indexed correctly, and existing stuck products were backfilled.
Products with lengthy descriptions can now be included in the search data used by AI Agent. Descriptions that previously prevented a product from appearing in AI search are handled without leaving that product out of the available catalog information.
Your team can keep detailed product descriptions in Shopify without those products being omitted solely because the description is too long to process. This fixes discovery of affected products; it does not guarantee that every detail in a long description will appear in an answer.
Keep AI Product Data synced and check representative product questions after updating your catalog. The change applies when product data is synced and does not need a separate chatbot option.
Related guide: What Your AI Agent Can and Can’t Help With.
Long email conversations with many attachments could take 30 to 60 seconds to open, and sometimes failed with an error. They now open normally.
Your team can open an email conversation with a long history and several attachments to review it and continue replying. The change addresses opening those larger threads; it does not require removing earlier messages or attachments from the conversation.
Select the email ticket from your inbox and wait for the thread to load. Review the messages and attachments, then use the reply box as usual. There is no new account setting to enable for this improvement.
Related guide: Onboarding: Understanding the Support Inbox.
In AI Chatbot settings, under Business Details, you can now force every reply into one language instead of mirroring the customer. Choose "Reply in customer's language" (the default) or "Always reply in a set language."
You can now choose whether AI Agent mirrors the customer’s language or consistently uses a language you select. This gives a channel a fixed reply language when that matches your support process or the language your team uses.
The language choice sits under Business Details and applies to the selected support channel. A fixed reply language is a separate choice from a fallback language used only when the customer’s language is unclear.
Open AI Agent settings, select the channel, and go to Business Details. Choose Always reply in a set language, select the language, and save. Choose Reply in customer’s language if you want the bot to mirror customers again.
The Facebook connect flow now loads your page list without waiting on one slow request, so the screen no longer spins forever. If something does go wrong, it shows the real error instead of failing silently.
When connecting a Facebook account with several pages, the page-selection step can finish loading without waiting indefinitely on one slow request. If the connection cannot continue, the flow shows an error that your team can use to decide the next step.
Open Settings → Support Channels, add the Facebook/Instagram connection, and complete the Facebook login and page selection. Choose the pages your team needs to manage and finish the permissions steps. The loading change is part of that flow and requires no separate setup.
Related guide: How to integrate Facebook with DelightChat?.
Sending a WhatsApp message with a missing or broken dynamic CTA button link used to crash the whole send with no record. The message is now marked "Undelivered" so you can see it failed instead of it vanishing.
If a required dynamic website-button value is missing, the attempted send remains visible as an Undelivered result. Your team can now find that failed attempt in performance reporting instead of losing the record. The message is not sent with a broken destination.
Review Undelivered results in the relevant automation's Performance tab. Check the template's website-button configuration and the value selected for its dynamic destination. For a broadcast, also check that the audience data supplies the variables the campaign uses. Correct the missing value before trying the message again. The broadcast guide's troubleshooting section covers required variables, valid CTA URLs and inspecting delivery failures. The failure record is created automatically; no separate setting is needed.
A ready export shows a countdown like "(Expires in 5 days)" so you know to download it before the 30-day window closes. Once it ages out, it consistently reads "expired" with a note to contact support for a re-run.
A completed export shows how long remains to download the file. Your team can use this notice to collect the data before the link is no longer available, rather than discovering the expiry when opening an older request.
Open Reporting → Conversation Export and check the completed request in the history table. Download it while it is available. If the row says expired, contact support for help with a new export of the same period.
The WhatsApp Business API self-serve setup wizard has a cleaner layout and copy, an accurate step count, and tidier translated instructions.
The setup wizard presents the WhatsApp API connection steps with clearer labels and a step count that matches the flow. It also improves the translated instructions, so teams can follow the setup in their preferred language. The change helps you see what information to provide and which step comes next.
Open WhatsApp API setup from Support Channels and follow the wizard for your connection flow. Review the information shown at each step before continuing. Existing connected numbers do not need to be reconnected just to use the clearer interface. For the rest of your workspace setup, see the DelightChat account setup guide.
Each of the 10 custom guardrail inputs in AI Chatbot settings now shows a different, relevant example instead of repeating the same one, making it easier to write a good rule.
Each custom-guardrail row now gives you a different example to help describe a situation that should be handled by your team. You can use those prompts to distinguish several business exceptions instead of starting every rule from the same example.
The examples are placeholders. They do not become active rules unless you enter and save your own instruction. Write the situation you want detected clearly, and keep each rule focused on when to hand over to a human.
In AI Agent settings, choose the channel and open Custom Guardrails. Add a row, replace its example with your own rule, and save. Existing saved rules remain your instructions.
Related guide: Set Up and Enable Your AI Agent.
If a search hits a database timeout, it now retries automatically a couple of seconds later, so a search that would have failed gets a second chance to succeed.
If an inbox search encounters a temporary database timeout, the app makes another attempt automatically. Your team can keep using the same search instead of restarting the lookup immediately after the first timeout.
Search for a customer's ticket from the inbox as usual. The retry runs in the existing search flow without a new setting. If a search still fails, check the displayed error and try again rather than assuming the ticket does not exist.
Related guide: Onboarding: Understanding the Support Inbox.
In AI Chatbot settings you can now add up to 10 custom rules that always hand a chat to a human, for example "payment proof screenshot." When a chat matches a rule, the bot escalates it instead of replying itself.
Some requests should always reach a person even when AI Agent can answer the surrounding topic. You can now write channel-specific guardrails that describe those exceptions, such as a request involving payment proof or a case that needs manual approval.
When the chatbot matches a custom guardrail, it escalates the conversation instead of replying on its own. These rules add your business-specific boundaries alongside the chatbot’s existing scope and safety rules; they are handoff instructions, rather than prompts that make AI perform a restricted action.
Open AI Agent settings, select the channel, and open Custom Guardrails. Add a clear description of each situation that needs a person, then save. Check Auto Transfer so matched conversations reach the intended team.
Related guide: Set Up and Enable Your AI Agent.
When the AI chatbot escalated a conversation to a human, the assigned agent could get cleared. The human assignee now stays assigned after the hand-off.
A conversation handed from AI Agent to your team should keep a human owner once one is assigned. The handoff path now applies the assignment rule without losing the human assignment during the transition.
The receiving agent can continue handling the ticket and read the private escalation note explaining why AI stepped back. This supports an accountable handoff rather than leaving the conversation without the intended owner.
Review the selected channel’s ticket assignment rule under Auto Transfer in AI Agent settings. Choose the team, team member, or unassigned destination that fits your support process, then save.
Related guide: Work Alongside Your AI Agent.
Templates now allow up to 10 quick-reply buttons (was 3) and two "Visit Website" buttons at once. This matches WhatsApp's current limits and gives more flexible layouts.
Give customers more reply choices or two different website destinations within a message template. For example, a quick-reply layout can offer several support topics, while a website-button layout can point to both an offer page and a product page. The additional options let you shape the message around the action you want the customer to take.
Open Message Templates and create a template. In the Buttons section, choose Quick Replies or Call to Action, then add and label the buttons for your intended layout. For website buttons, configure the corresponding destinations when setting up the broadcast or automation that uses the template. Submit the new template for approval before sending it. The editor's controls show the available button combinations.
Uploading a CSV for a broadcast or segment now flags rows that will be skipped or fail, such as bad phone numbers or duplicate country codes. You can fix problem rows before sending instead of finding out afterward.
Review audience problems as soon as you upload a CSV. The upload result lists warnings alongside the accepted contact count, so your team can correct formatting problems before using the file in a campaign. This is useful when a customer export contains phone numbers in a different format from the broadcast uploader.
Upload the audience CSV while creating a broadcast or a CSV-based customer segment. Read any warnings, correct the affected rows in your spreadsheet, and upload the corrected file again before sending. Keep Country Code and Phone Number in the separate required columns. The Shopify export formatting guide shows how to convert combined phone numbers into that format.
Outbound WhatsApp template messages are now part of the history the AI chatbot and Generate with AI read. Replies are more context-aware when a template was sent earlier in the chat.
A shopper’s reply to a WhatsApp template often makes sense only alongside the message they received. AI Agent and manual AI reply generation now include relevant outbound templates in the recent conversation history they read.
For example, an earlier order-confirmation or shipping template can provide context for the customer’s next question. Templates participate in the recent-history window like other messages; this does not mean every old campaign remains in the chatbot’s context forever.
Continue sending WhatsApp templates through your existing workflow. Use Generate AI Reply in the inbox for a manual draft, or configure automatic replies in AI Agent settings. No separate template-history switch is required.
Related guide: AI Assistant Settings (Manual).
Typing "1. " or "- " in a WhatsApp, Instagram or Facebook reply had stopped turning into a list. Numbered and bulleted lists auto-format again.
You can write steps or multiple points as a list directly in a chat reply. The editor recognizes the list prefix while you type, keeping the reply organized without requiring you to format each line manually.
In a WhatsApp, Instagram, or Facebook reply, begin a line with 1. followed by a space for a numbered list, or a hyphen followed by a space for a bulleted list. Continue writing and review the formatting before sending.
Related guide: Onboarding: Understanding the Support Inbox.
How many past messages the AI chatbot considers is no longer one account-wide setting. Admins can now tune it per channel.
Different support channels can need different amounts of recent conversation context. The number of past messages supplied to AI Agent is now stored per channel, rather than taken from one account-wide setting.
Changing one channel’s context setting therefore does not require using the same value for the other channels. This controls the recent messages considered for an answer, rather than creating permanent or unlimited memory of every conversation.
Channel-level context uses the existing AI Agent channel configuration. The current published help guides do not document a memory control; contact your DelightChat support team if you need help checking or changing this setting.
A new Conversation Export screen under Reporting lets an admin pick a channel and date range and request an export. You get a secure download link when the export is ready, with no need to email support.
Admins can request a copy of conversation history directly from the dashboard and follow its status in one place. The export is prepared in the background, so submitting the request does not produce an immediate download.
The original launch had a shorter date-range limit. A later update lets you select a longer period and splits it into separate exports. Follow the current screen for the applicable limits and expected wait.
The chatbot now considers multiple product questions across recent conversation turns, instead of looking up only the latest product mentioned. This helps it continue a conversation about more than one item.
A customer may ask about a second product while an earlier product question is still unresolved. AI Agent now uses the pending product questions in recent conversation context to retrieve information for more than the latest product mentioned.
This helps the chatbot answer multi-product catalog questions without dropping an earlier item simply because the conversation moved to another one. Recent context remains bounded, and branch-specific inventory checks follow their own product-and-location flow.
No new memory toggle is needed for this retrieval improvement. Keep AI Product Data current, and use full product names when discussing several items.
Related guide: What Your AI Agent Can and Can’t Help With.
Merchants can now connect a WhatsApp Business number through a guided, step-by-step wizard instead of a support-led setup. It walks you through Business Manager linking, Gupshup setup and Meta go-live, with in-wizard help at each step.
The guided connection flow brings the required account and number setup into the app. You can work through the steps in order and use the help shown at each stage, rather than relying on a separate support-led setup for the whole process.
Go to Settings, open Support Channels, and choose Add a WhatsApp API channel. Follow the wizard's instructions for your business account and phone number, then complete the verification steps it presents. External account approval and number verification are still required.
Related guide: Onboarding: Setup your DelightChat Account.
Product search gives product names and word matches more weight when distinguishing similarly named items. The chatbot can use a shopper's specific wording to find the relevant product.
Catalogs often contain products with almost the same names. The chatbot’s product retrieval gives word and product-name matches more weight, so a shopper’s specific name or model wording can help distinguish closely related items.
This improvement supports product conversations where a broad semantic match is not enough. It does not make missing catalog information certain: a useful answer still depends on identifying the relevant product and using the store’s available details.
Keep product titles and descriptions clear in Shopify and keep AI Product Data synced. Customers can include the full product name, variant, or model wording in their question.
Related guide: What Your AI Agent Can and Can’t Help With.
The chatbot's stock answers now honor Shopify's continue-selling setting for each product variant, including location availability answers. It distinguishes availability to order from a confirmed physical stock count.
An item that Shopify allows customers to order after its tracked quantity reaches zero should not automatically receive an “unavailable” answer. AI Agent now distinguishes availability to order from a confirmed physical stock count.
For a variant with Shopify’s continue-selling policy, the chatbot can say it is available to order when the product data supports that. It should not invent a positive unit count at a branch whose physical stock is unknown or depleted. Your Shopify selling policy still determines whether the item can be purchased.
Maintain the variant’s selling policy in Shopify. If you use branch-level stock answers, keep Share location-wise availability enabled for the channel in AI Agent settings. This change adds no separate continue-selling setting in DelightChat.
Shopify is retiring the script this banner relied on, so it had already stopped displaying. Its nav links and help text are now hidden. The COD-to-Prepaid automation itself is unaffected.
The removed links belonged to the old banner on Shopify's Thank You page. Removing them stops sending your team to instructions for a storefront feature that no longer displays. This change does not turn off your WhatsApp COD-to-Prepaid automation or change its configured order-confirmation rules.
Manage the WhatsApp flow under Marketing → Automations. It can send eligible COD customers a payment link according to your existing automation settings. You do not need to enable a replacement banner for that message flow. Use the COD-to-Prepaid automation guide to review the payment provider, message template and follow-on order actions; the retired banner guide is not the setup guide for this workflow.
Generate AI Reply now streams the draft into the reply box word by word instead of all at once. Send stays disabled until the draft finishes.
You can now see a generated reply take shape inside the inbox composer while AI is still writing. The reply appears progressively, which lets your agent see that generation is underway instead of waiting for the entire draft to arrive at once.
Send and Send & Close stay disabled while the draft is streaming. When generation finishes, the agent can read the complete response, make changes, and send it. A visible partial draft is not a message already sent to the customer.
Click Generate AI Reply in the reply box, wait for writing to finish, then review and send. Configure manual response generation in AI Assistant settings.
Related guide: AI Assistant Settings (Manual).
A small Give Feedback button now sits beside the heading on every dashboard page. Anyone can leave quick feedback about a page without leaving it.
Page feedback gives your team a way to explain what worked or what was difficult while the screen is still in front of them. Feedback is associated with the feature you are using, which provides more context than an unrelated general suggestion.
Use the feedback controls near a page heading and add a comment when prompted. Describe the action you were trying to complete and the result you expected. You can continue your work on the same page after submitting.
The Automations screen has a new Library tab and Browse Templates button. You can browse and copy ready-made automation flows instead of building each one from scratch.
Find an automation example before starting a new flow. The library brings ready-made flows into the Automations screen, so your team can browse an existing starting point and adapt it to your store's goal. The Library tab is available for channels linked to a Shopify store.
Open Automations for the relevant channel, then select Library or Browse Templates. Review the selected example and follow its instructions to copy or set up the flow. Check the trigger, conditions and approved message template before enabling it. The Shopify-trigger automation guide explains the configuration, enable switch and save steps.
The chatbot sometimes offered to check other branches before answering the shopper's original question. It now answers the current location first, then offers to check elsewhere.
A shopper asking about one store location should first get an answer about that location. AI Agent now puts the requested availability answer before an offer to check another branch, so the shopper can use the first part of the reply to decide what to do next.
When another branch may help, the chatbot can offer that as a follow-up. It uses the named location and product data available to it; it does not need to turn every stock question into an extra branch-selection step.
For branch-level answers, check Share location-wise availability in the selected channel’s AI Agent settings. Keep the connected Shopify location and variant inventory accurate.
When a shopper mentioned an order number, the chatbot could look it up without confirming ownership. It now verifies the order belongs to that shopper before sharing details.
An order number by itself is not enough to show that an order belongs to the person asking about it. AI Agent now checks ownership before it shares details for a customer-supplied order number.
If ownership is not yet confirmed, the chatbot asks for the email address or phone number used for the order. It does not disclose the order’s contents, status, tracking, or totals while that check is unresolved. Customers can supply either of those identifiers to continue the lookup.
The check runs in the existing order-answering flow. Your team does not need to add a custom guardrail for it. Review the enabled order capabilities and human handoff destination in AI Agent settings.
Related guide: What Your AI Agent Can and Can’t Help With.
The chatbot sometimes told shoppers a product didn't come in a certain size or color without actually knowing. It no longer invents these denials when unsure.
A missing product detail is different from proof that an option does not exist. AI Agent now avoids turning absent evidence into a categorical statement that a product never comes in a size, color, or other attribute.
The chatbot should ground its answer in the product information it has, ask for a useful clarification when needed, or hand over when it cannot answer. This helps shoppers distinguish a confirmed product fact from information that still needs your team to check.
Keep Shopify product details and relevant AI Knowledge Base answers accurate. The evidence rule applies to the existing AI Agent flow; there is no separate product-detail switch to enable.
Related guide: What Your AI Agent Can and Can’t Help With.
Sharing a product search link sometimes triggered an unnecessary hand-off to a human. Search links are now treated as a normal informational reply.
A store search link can help a shopper browse products without requiring a human to intervene. AI Agent now handles store search and other supported browsing links as informational context instead of treating the link alone as a reason to escalate.
This keeps a normal product-browsing conversation on the chatbot’s answering path. Requests that genuinely need your team still follow the existing handoff rules; sharing a browsing link does not override those rules.
No separate setup is required for this link handling. Continue to maintain the connected store and the selected channel’s product capabilities in AI Agent settings.
Stock answers for a named store location didn't account for which size or color the shopper asked about. They now reflect the specific variant.
A branch can have one size or color available and another sold out. AI Agent now keeps that difference in the answer instead of using one variant’s inventory as proof that every option is available at the location.
If a shopper has named a variant, the answer checks that variant. If variants differ and the shopper has not chosen one, the chatbot can state which options are available and which are not. A product-level “available” answer is not used to hide those differences.
Use Share location-wise availability in the channel’s AI Agent settings, and keep variant inventory correct in Shopify. The variant-aware answering rule applies within that stock flow.
Editing an FAQ in the AI knowledge base didn't refresh what the chatbot searches, so it kept using the old answer. Edits now update the search index immediately.
Changing a policy or answer in the AI Knowledge Base should also change the information the chatbot can retrieve. Saving an existing FAQ now refreshes its question and answer in the search data used for AI replies.
Your team can maintain one approved FAQ instead of adding duplicate answers to make an edit take effect. This matters when the content changes, such as a revised policy or business process. The chatbot still needs a relevant approved answer to ground its response.
Open the AI Knowledge Base, choose the brand if needed, edit the FAQ, and save it. Keep the final text factual and check a related customer question after the update.
Related guide: AI Knowledge Base - Create, manage, and organize helpful articles to provide quick answers to common queries..
The monthly analytics report used a stale date window, so numbers didn't match the month shown. It now reports for the correct date range.
The monthly email report uses the reporting month's dates when calculating your marketing results. Its date window follows the brand's time zone, so the totals describe the month being reported rather than an earlier period. This helps your team review the same period when comparing the email with marketing analytics.
Open Marketing Analytics Settings to review whether monthly report emails are enabled and who receives them. Existing enabled reports use the corrected calculation automatically. When comparing a report with Overview Analytics, use the dates shown in the email and the same channel. You do not need to recreate your automations or broadcasts.
The template editor now allows up to 1000 characters in the body (was 550), minus 50 for each variable used. The character counter shows exactly how much room your variables take.
There is more room to explain an offer, an order update or the next step in a WhatsApp template. The editor also accounts for space reserved for variables, so adding personalized fields changes the available space shown by the counter. This lets you refine the message in one place before submitting it for approval.
Open Message Templates, create a template and write the message body. Add the variables you need and use the counter in the editor to check the remaining space as you write. Submit the completed template for approval before using it in a broadcast or automation. See the template creation guide for naming, categories, variables and submission.
Most Marketing and Settings pages now show a "See Guide" button next to the heading that opens the relevant how-to guide instantly. Guides used to live only in a separate tab that was easy to miss.
When you need help with a setting or marketing feature, you can open that page's guide from its heading. This keeps the instructions close to the work you are doing, rather than requiring you to search through a separate help section first.
Look for the guide button beside a supported page's heading and open it to read the instructions. The guide appears in the existing help drawer. You may see the label How it works on the current interface; no setting is needed to make guides available.
Generate AI Reply always quoted prices in USD and could link to a placeholder storefront. Drafts now quote your actual currency and link to your own store.
A product price needs its currency, and a useful product link needs to lead to your storefront. Generate AI Reply now gives the drafting system Shopify’s product or store currency and constructs variant links from your connected store’s domain.
Your team can review a draft containing product prices and links without first correcting a default USD assumption or replacing a placeholder storefront link. The reply remains a draft for an agent to check before sending; this change does not turn manual drafting into an automatic customer response.
Use Generate AI Reply in the inbox reply box. For product-based drafts, review the product-data options in AI Assistant settings and your connected Shopify product data.
Related guide: AI Assistant Settings (Manual).
The chatbot previously assumed every merchant had a connected Shopify store, so non-ecommerce channels couldn't fully use it. Merchants without an online store can now use it, with prompts suited to non-ecommerce conversations.
Businesses without an active Shopify store can use AI Agent to answer questions from their FAQs and business details. The chatbot uses a non-store conversation path instead of assuming it can look up products and orders in Shopify.
This is useful for questions about your services, business hours, or policies that you have supplied. Requests that need store data or a transaction still need a human when the chatbot cannot answer from that information; connecting no store does not create an order-tracking or inventory source.
Add approved answers in the AI Knowledge Base, then configure the channel in AI Agent settings. Add accurate business details, choose the available capabilities and handoff settings, and save.
Related guide: AI Knowledge Base - Create, manage, and organize helpful articles to provide quick answers to common queries..
When a customer's last message was mostly a link, emoji or contact card, the chatbot could hand off in the wrong language. Hand-offs now fall back to your configured default language when there's too little text to detect one.
A link, an emoji, or a shared contact card may not contain enough natural-language text to identify the customer’s language. AI Agent now uses the channel’s configured fallback language for this situation instead of guessing from a URL or another non-sentence fragment.
The same fallback is used when preparing a handoff message for the customer. This gives you a language to use when the latest input cannot establish one, while ordinary language-bearing messages still follow the chatbot’s language settings.
In AI Agent settings, choose the channel and open Business Details. Review the fallback language and save the setting. Review the handoff-message text in Auto Transfer as well.
Related guide: Set Up and Enable Your AI Agent.
Turn on "Share inventory count" and "Share location-wise availability" in AI Agent settings, and the chatbot can tell a shopper the exact stock count or whether an item is available at a specific store. Previously it could only say in stock or not.
You can choose how much inventory information AI Agent shares when a shopper asks about availability. Stock counts answer “how many are left?”, while location-wise availability helps a customer check a particular store or branch.
These are separate sharing choices. If counts are off, the chatbot should not quote a unit count. If location sharing is off, it should not give branch-level stock information. Numbers and locations come from the Shopify data available to the chatbot, rather than estimates.
Open AI Agent settings for the channel. Under the product-question capability, turn on Share inventory count, Share location-wise availability, or both, then save. If a Shopify reconnection notice appears, complete that step so the required inventory data is available.
A broadcast whose send time had already passed when you published was created but never sent, with no warning. Publishing now validates the time and shows a clear error.
When you publish a scheduled campaign, DelightChat checks that its chosen send time is still valid. If the time has passed while you were preparing the campaign, you get an error so you can correct the schedule. This makes the next action clear instead of leaving a campaign that looks ready but never starts.
Create or edit a campaign in WhatsApp Broadcasts. Before publishing, review the date and time under message scheduling. If the app shows a scheduling error, choose a future time and publish again, or select the immediate-send option if that is your intent. The validation applies automatically. For the complete audience, template and scheduling flow, see the broadcast setup guide.
The "Share Product Feedback" page no longer uses a third-party embedded form. You can submit feedback (with up to 3 attachments) directly in the app, with formatting tools and a confirmation screen.
You can describe a product suggestion and add supporting files from a form inside DelightChat. Formatting lets you explain the situation clearly, and the confirmation screen tells you that the feedback was submitted.
Open Share Product Feedback, explain what you were trying to do and what would help, and attach an example if needed. Submit the form and check the confirmation before leaving the page.
The drawer showed generic text like "New reply from customer…". It now shows the customer's name, what happened, the channel, and a message preview, with relative timestamps like "5m ago".
Your team can see which customer and channel an update concerns before opening the conversation. The message preview and description provide context for the activity, while relative times help you identify recent updates in the drawer.
Open notifications from the bell icon in the left navigation and select an update to review its conversation. The richer information appears in the existing drawer; you do not need to enable a new notification type.
Related guide: Where can I view all the notifications?.
Add Fastrr / Shiprocket SR Checkout as a checkout provider under Settings → Integrations, with its own name and icon.
Stores using Shiprocket Checkout / Fastrr can identify and manage that provider as its own integration in DelightChat. This keeps the checkout provider separate from the Shopify store connection and gives your team the setup instructions for the selected provider.
Open Settings → Integrations, add a checkout provider, and choose Shiprocket Checkout / Fastrr. Select the store and follow the provider-specific instructions shown on the setup page. Complete those steps before enabling and testing the recovery messages you want to send.
Related guide: How to send Automated Abandoned Checkout Recovery messages via WhatsApp.
Overview, Automation and Broadcast analytics now show a graph of messages attempted, sent, delivered and read. Channels without a connected store can now view message analytics too (revenue figures stay hidden for them).
See the stages of message delivery together instead of comparing separate totals. The graph shows how many messages were attempted, sent, delivered and read over the selected dates. Use it to spot a change in campaign performance and decide which delivery stage to examine in more detail.
Open Marketing Overview Analytics, Automation Analytics, or Broadcast Analytics, then select the channel and date range you want to review. The graph appears in the Analytics view without a separate switch. Channels without a connected store can still review message performance; store-based revenue figures are hidden. The broadcast guide's analytics section explains the delivery results and downloading campaign data.
Recovery amounts were shown in paise instead of rupees, making them look 100 times too large. They now show the correct value.
New Razorpay Magic recovery messages use the checkout amount in rupees instead of treating paise as rupees. The same conversion is used for new recovery revenue, so a checkout amount is not multiplied by one hundred in the message or its newly recorded revenue.
Your existing recovery automation continues to run without a new setup step. Review new messages and their results in WhatsApp Marketing Automations. This was a forward-looking correction: it did not rewrite older recovery records or historical revenue figures.
Related guide: How to send Automated Abandoned Checkout Recovery messages via WhatsApp.
AI settings were grouped under confusing names like "AI Chatbot Settings" and "AI Global Settings". They're now separated into AI Agent (Chatbot), AI Assistant (Manual), AI Product Data, and AI Knowledge Base.
The AI section now separates the chatbot that answers customers automatically from the assistant your team uses while writing replies. Product data and knowledge-base content have their own destinations, so you can open the setting that matches the task you are doing.
Use AI Agent (Chatbot) for automated replies and handoff behavior. Use AI Assistant (Manual) for tools used by your agents, AI Product Data for Shopify product sync, and AI Knowledge Base for approved FAQs.
Open Settings and expand the AI section. Go directly to AI Agent, AI Assistant, AI Product Data, or AI Knowledge Base.
Related guide: AI Assistant Settings (Manual).
App files (JS/CSS) were served from US servers for everyone. They now load from a nearby CDN, so the app loads faster outside the US with no change elsewhere.
The web app can load its page files from a nearby delivery location rather than requiring every visitor to download them from the United States. This change is relevant to teams working from other regions when they open the dashboard.
Open DelightChat and sign in as usual. The app uses the updated delivery setup automatically; there is no account setting, installation, or change to your support workflow.
Agents could type comment replies longer than Instagram, Facebook or Messenger accept. A live counter now prevents sending a reply over each platform's limit (Messenger 1000, Instagram 2000, Facebook 5000).
The reply editor shows how much room remains for a response on the channel you are using. If a reply goes over that channel's limit, you can shorten it before sending rather than discovering the limit after the send attempt.
Open the customer's conversation, write your reply, and check the counter beside the editor. Keep the reply within the displayed limit. This works in the reply flow; your team does not need to enable a separate setting.
Related guide: Facebook & Instagram Comment Support in DelightChat.
A key-matching bug meant recovery attribution never linked back to the order, so tracking undercounted recovered sales. Recovered Razorpay Magic checkouts now link back correctly.
When a customer completes a Razorpay Magic checkout after a recovery message, the recovery can now be linked to the order. Your team can review the resulting recovery attribution rather than treating an unlinked order as an unsuccessful follow-up.
Open your abandoned-checkout recovery automation in WhatsApp Marketing Automations and review its analytics. The correction applies to the existing recovery flow; you do not need to rebuild the automation to use it.
Related guide: How to send Automated Abandoned Checkout Recovery messages via WhatsApp.
Drafts weren't always saved or deleted, so clearing a reply box could leave a stale draft. Drafts now save as you type and clear properly, across chat and email.
A reply you have started stays available as a draft while you work on the conversation. Clearing the reply box also clears the saved draft, so returning to the ticket should not bring back text you have already removed.
Write in the reply box as usual, then return to the conversation when you are ready to continue. If you want to discard the reply, clear the composer. The draft behavior applies to chat and email replies without a separate setup step.
Related guide: Onboarding: Understanding the Support Inbox.
Searching an email address while composing used a slow, unindexed lookup. It now returns the top 20 matches faster.
Recipient suggestions appear while you enter an email address in the compose window. The improved lookup helps you find an existing contact without typing the complete address, especially when several people have similar names or addresses.
Open Compose, select your email channel, and start entering the recipient's address. Choose the matching suggestion, then check the recipient before sending. Your team can use the same compose flow; there is no new setting to turn on.
Related guide: How to compose a new email?.
Open ticket counts in the inbox now update in real time across channels, teammates, and tags. When a ticket is assigned, tagged, closed, or reopened, the related counts update automatically without needing to refresh the page.
This makes the inbox more accurate and reliable during active support workflows, especially for busy teams managing high ticket volumes. The counts now stay in sync across open tabs, helping agents quickly understand workload, ownership, and ticket status while they work.
Mobile push notifications now include more helpful information for every new customer message. Each notification shows the customer's name or phone number, the support channel name, and the customer's message body where available, so your team can quickly understand who is messaging and what the conversation is about.
When both a name and phone number are available, the notification displays the customer's name. If only a phone number is available, that number is shown instead. By including the message preview directly in the notification, agents can better prioritize incoming conversations and respond faster from the mobile app.
You can now request an export of your DelightChat chat conversations by contacting our support team from here.
When raising the request, please mention the support channel and the date range you want to export. Once requested, you'll receive the export within 24 working hours via email, with a downloadable CSV file.
The export includes the conversation messages exchanged between customers and your team, including the back-and-forth message content for the selected channel and period. This can be useful for internal reviews, support quality analysis, reporting, and AI analysis of real customer conversations - such as identifying common questions, improving help content, and generating AI Knowledge Base FAQs from actual customer interactions.
When the AI chatbot encounters a customer question it cannot confidently answer, the query is now automatically captured and added to the AI Knowledge Base review section. This gives teams a clear list of unanswered questions that need attention, instead of relying on manual tracking or agent feedback.
Each entry includes the customer's question along with relevant conversation context, making it easier to understand what the customer was asking and create the right Knowledge Base content. Over time, this helps you continuously train and improve the chatbot using real customer conversations, reduce repeated unanswered queries, and improve AI response accuracy.
The latest DelightChat mobile app release includes several improvements to make the app faster, smoother, and more reliable. Version 3.0.2 brings stability enhancements, faster load times across key screens, smoother transitions between tabs, improved notification handling, and refinements to email cards for a cleaner mobile support experience.
We've also added a new gesture-based navigation improvement: you can now swipe from left to right to go back to the previous screen in the app. This makes it easier to move between conversations, customer details, and other screens while managing support on mobile.
DelightChat now supports direct webhook integrations with custom checkout providers such as FlexyPe, GoKwik, Razorpay Magic Checkout, Cashfree / Ecom360, Cashfree / OCC, and Shopflo for abandoned checkout recovery. These integrations help merchants capture checkout events directly from their checkout provider without relying on intermediary connectors.
Once connected, DelightChat can listen for checkout activity from the selected provider and trigger abandoned checkout recovery workflows more reliably. You can set this up from Settings → Integrations → Add Integration in your DelightChat dashboard.
You can now view message reactions and read receipts in Facebook Messenger and Instagram Direct Message conversations inside DelightChat. Reactions appear alongside messages in the conversation thread, giving agents more context on how customers are responding.
Read receipts also show when your message has been seen by the recipient. This gives support teams better visibility into customer engagement and helps agents decide when to follow up at the right time.
DelightChat's new AI Chatbot is now available for Shopify merchants across WhatsApp, Instagram DMs, Facebook Messenger, and Live Chat. The chatbot can answer customer queries with richer context from your store, including customer details, product information, order data, and your AI Knowledge Base.
The AI works alongside your existing inbox workflows and is designed to choose the right next step in a conversation - whether that means answering directly, asking a clarifying question, or escalating to a human agent. You can configure strict guardrails from chatbot settings to decide what the AI should and should not answer, along with escalation and handoff scenarios for cases where your team should take over.
The chatbot also supports smarter product matching and identification. It can understand customer intent from product images, shared product links, partial names, jumbled text, or unclear descriptions, and match them to relevant products from your Shopify store. This helps the AI handle common ecommerce questions such as order status, product availability, return policies, delivery timelines, product recommendations, and FAQs more effectively - reducing repetitive workload for agents while giving customers faster answers. Get started setting up your AI agent from here.
DelightChat now supports better handling of Partially Paid and Partial COD orders, especially for brands using custom checkout providers where these payment statuses are commonly used.
When an order is only partially paid, you can now trigger automations based on the order's payment status and include payment-specific details in your WhatsApp messages.
We've added two new message template variables:
{AmountPaid} — shows the amount already paid by the customer{AmountDue} — shows the remaining balance due for the order
To set this up, go to Marketing → Automations → Create New Automation. Select Shopify Triggers, then choose New Prepaid Order as the event trigger.
In the entry conditions, set: Payment status of order → Partially paid
Once this condition is added, choose your message template and use {AmountPaid} and {AmountDue} to clearly communicate the paid amount and pending amount to the customer.
A new toggle in Team Member settings labeled "Mask Customer Phone Numbers in Support Inbox" allows you to control the visibility of sensitive information. When enabled, customer phone numbers are masked in the web and mobile support inbox views—showing only the country code and last two digits, with the remaining digits replaced by "xxxx".
Read receipts are now available for WhatsApp session messages in DelightChat. You can see when a message is sent, delivered, and read directly within the conversation thread, giving agents better visibility into message status during customer interactions.
Message status is indicated using familiar WhatsApp indicators: a single tick means the message was sent, double ticks mean it was delivered, and double green ticks mean it was read. Read confirmations will only appear if the recipient has read receipts enabled on their WhatsApp account. This helps teams better understand customer engagement and follow up at the right time.
This will be released in a phased manner to all accounts within the next few days.
You can now import knowledge base content from multiple sources into the AI Knowledge Base.
This update allows you to connect, upload, and consolidate information from supported platforms and file types into a single knowledge base - giving you greater flexibility and control over how you manage your content.
Bulk Import enables you to quickly add multiple FAQs to your AI Knowledge Base. In both options below, you'll be able to review all FAQs before they are added.
Import from URL
Provide a URL and AI will automatically scan the page and extract relevant FAQs. You can review and edit the imported FAQs before adding them to your Knowledge Base.
Upload a CSV
Upload a CSV file containing question–answer pairs. The file must include two columns: question and answer. You'll be able to review and confirm the FAQs before publishing them to your Knowledge Base.
You can now send dynamic media through the custom API (via external triggers) automation for templates with video and document headers. This extends the previous support, which was limited to image headers, allowing custom API-triggered messages to dynamically reference external video or document resources in template headers.
AI-generated replies were primarily based on text context, which can limit accuracy when customers share images or voice notes.
Going forward, image and voice context in AI-generated responses is enabled - allowing the AI to better understand customer intent from richer inputs. This will help produce more relevant, accurate replies and improve overall conversation quality.
Several interface elements on the mobile app have been updated for clarity and consistency. Labels are now more readable in low-light mode, touch targets for key actions have been expanded, and notifications adopt your device's accessibility settings for text size. Navigation transitions have also been refined for a smoother experience across devices.
AI now has context of Shopify product catalog. When drafting replies, AI pulls in contextual details like product options, titles, prices, categories and links directly from your connected Shopify store. To sync latest Product data, go to AI settings.
You can now guide AI-generated replies using explicit agent-provided directional context. When generating an AI reply, any text added in the "Add AI Context" field is passed to the AI as directional input to better understand your intent and the conversation.
This context helps the AI align its response with what you want to communicate, without directly copying or repeating the input unless it's contextually appropriate. If no context is added, AI reply generation works as usual. This gives agents more control over tone, intent, and accuracy while still benefiting from AI assistance.
Audio messages now show their transcriptions directly inside the conversation.
You can now trigger DelightChat actions directly from Shopify Flow.
Quickly view and navigate between connected support tickets.
Access help and resources directly from the mobile app.
+countrycode-number / +countrycode-0-number / number / countrycode-number / 0-number / countrycode-0-number / +number)Cash-on-delivery orders are now automatically tagged based on WhatsApp verification status.
CODNoResponse — message delivered but no response receivedDC_COD_OptedOut — customer opted out of WhatsApp communicationDC_COD_NotReachableOnWhatsApp — customer couldn't be reached on WhatsAppGenerate complete replies in seconds with AI that understands your business context.
Reply to individual messages inside a conversation, just like the native WhatsApp app.
Use DelightChat in the language you're most comfortable with.
Send and receive WhatsApp voice notes directly from the DelightChat mobile apps.
Automatically retry failed broadcast messages caused by frequency limits.
Send rich, scrollable product carousels on WhatsApp.
"Edit message tone and clarity directly inside the editor."
"See product updates directly on login so teams never miss important changes."
Automatically convert Cash-on-Delivery orders to prepaid.
Show a payment banner on the Shopify thank-you page for COD orders with a Razorpay payment link.
Use Razorpay to collect prepaid payments for COD conversions. (More gateways planned)
New Paused and Disabled states for templates.
Assign, close, or update multiple tickets at once.
Receive weekly and monthly performance summaries via email.
Notify customers on WhatsApp when out-of-stock products are restocked.
Unified view of WhatsApp marketing and automation performance.
Customize your workspace:
Set up CTWA automations directly from DelightChat.
Sell directly through WhatsApp.
Automatically send CSAT surveys when tickets are closed.
