- Should you act now or can you wait?
- What is this AI knowledge base and what is it really for?
- How does the automatic store scan work?
- What actually happens if the knowledge base is incomplete or wrong?
- Three real-world scenarios: what changes for an Italian e-commerce store?
- Why everyone is talking about it now: is AI commerce really growing?
- Where to start: the concrete steps to get your knowledge base in order
- What tools does Shopify use to let AI agents talk to your store?
- What mistakes are most common in this migration?
- How do I know if my AI knowledge base is working well?
- Rules and risks to keep an eye on
- So, is it worth it for your company?
- How we see it at SHM
- Frequently Asked Questions
- Shopify has replaced the Knowledge Base app with a section built into the admin: the old app stops working on January 15, 2027.
- If your store had more than 1,000 facts in the app, you need to re-enter them by hand: no automatic transfer took place.
- The AI knowledge base is public and accessible to all AI agents (including third-party ones): what you publish really matters.
- In Italy, 13% of shoppers already use AI in the pre-purchase phase: having accurate information in your knowledge base isn't a technical detail, it's sales.
Out of 88 Shopify stores tested, only 2 were able to correctly answer an order cancellation question. If your AI knowledge base isn't up to date, your customers get wrong answers, and you don't even know it.
Should you act now or can you wait?
It depends on how many facts you had in the previous app. If you had more than 1,000, you need to act before January 15, 2027 or you'll lose that data. If you were under that limit, the transfer has already happened automatically.
| Your situation | Worth acting? | why |
|---|---|---|
| You used the app with more than 1,000 facts | Yes Right away | Your facts were NOT transferred: you need to re-enter them manually by Jan 15, 2027 |
| You used the app with fewer than 1,000 facts | Yes Check | The transfer is automatic, but you need to check that everything is published correctly |
| You didn't use the app | No urgent action | The new admin is already active with the basic policies; you can enrich it whenever you like |
| You have a custom storefront or an Agentic plan | Yes Caution | You can only add 'additional facts': the standard categories can't be edited |
| You have a password-protected storefront | Limited | The automatic scan doesn't work: the knowledge base only shows data entered manually in the admin |
The January 15, 2027 deadline is not negotiable: after that date, the old app will no longer be available. Shopify sent direct emails to merchants with more than 1,000 facts, but it's worth checking your spam folder and company communications too.
If you weren't using the app, the good news is that some information, such as store policies and contact details, is published automatically when it's created in the admin. The key is figuring out what's missing and what needs to be added by hand.
What is this AI knowledge base and what is it really for?
It's the instruction document your Shopify store gives to AI agents, meaning artificial assistants, so they can answer customer questions. Without it, the AI makes up answers or doesn't answer at all.
Imagine a store assistant who has never read the internal manual: they answer at random. The knowledge base is exactly that manual, written for the AI agents that talk to your customers on your behalf.
Shopify organizes it into ten categories: returns and warranties, payment methods, delivery times, second-hand purchases, and so on. There's also a free-form category, Additional Facts, where you can add any information not covered by the others.
The critical point: this knowledge base is public. Anyone with access to the Storefront MCP (the technical tool Shopify gives developers to query the store) can read it, including third-party apps and external AI platforms like Shopify's own Shop app.
| Category | Data source | Published automatically? |
|---|---|---|
| Store policies (returns, shipping…) | Shopify admin or storefront scan | Yes Yes, if created in the admin |
| Contact | Shopify admin | Yes Yes, automatic |
| Payment methods | Storefront scan or manual input | Not yet No, requires approval |
| Delivery times | Storefront scan or manual input | Not yet No, requires approval |
| Secondhand purchases | Manual input | Not yet No, requires approval |
| Additional Facts | Merchant's manual input only | Not yet No, always manual |
How does the automatic store scan work?
Shopify periodically scans your storefront and suggests updates to the knowledge base. You decide what to approve or reject before it reaches AI agents.
The automatic scan is useful because it picks up information already on your site, like hours, policies and descriptions, and turns it into structured facts for AI agents. But it isn't foolproof: some site changes get missed.
If you update your return terms or shipping times and the scan doesn't catch them, you can request a new manual scan. The catch: it can take up to 48 hours. For an e-commerce store with frequent promotions, that delay can lead to outdated AI answers.
- 1 · ShopifyScans your storefrontAutomatically detects public information: policies, contact details, terms of sale.
- 2 · ShopifySuggests updatesNewly detected facts stay 'unpublished' until the merchant decides.
- 3 · MerchantApproves or rejectsYou can publish everything in bulk, by category or one field at a time. You have full control.
- 4 · AI agentUses only published factsNo content written by Shopify reaches an AI agent before the merchant approves it.
- 5 · MerchantRequests a rescan if neededIf your site has changed and the scan missed the updates, you can force a new scan (up to 48h).
What actually happens if the knowledge base is incomplete or wrong?
AI agents give wrong information or don't answer at all. A test on 88 Shopify stores found that only 2 could handle an order cancellation question.
The finding comes from a test of 122 Shopify stores, published on Search Engine Journal: of the 88 stores asked how to cancel an order, only 2 gave a usable answer. The problem? The standard knowledge base has no dedicated category for cancellations. It has to be added as an additional fact.
This isn't an abstract technical problem. If a customer asks your store's AI assistant how to cancel an order and gets no answer, or worse, a wrong one, the result is an avoidable abandoned cart or refund request.
The practical advice is simple: test your store with the real questions customers ask. Don't rely only on the query log from the previous app, because the new admin doesn't yet have an equivalent section for store questions.
Three real-world scenarios: what changes for an Italian e-commerce store?
The impact varies a lot depending on the size of your information catalog and how you used the app before. Here are three illustrative cases.
Illustrative examples: the numbers and situations are meant to show the reasoning, they are not real customer data.
| Store type | Typical situation | Action needed | Risk if you don't act |
|---|---|---|---|
| Fashion e-commerce (200 SKUs) | Used the app with ~300 facts on returns, sizing, shipping | Check that the facts were transferred and published correctly | Incomplete AI answers on return policies → more returns |
| Handicraft store (50 products) | Didn't use the app, policies are on the website | No urgent action; enrich with additional facts on customization and lead times | AI agent can't answer questions about custom production times |
| Multi-brand marketplace (5,000+ SKUs) | Used the app with 1,500+ facts on brands, policies, promotions | Manually re-enter facts beyond the 1,000 limit by Jan 15, 2027 | Loss of critical information → AI agent misinformed about your brand and promotions |
Why everyone is talking about it now: is AI commerce really growing?
Yes, and the numbers are concrete. In Italy, more than 13 in 100 online shoppers already use AI in the pre-purchase phase, and AI traffic to Shopify stores has grown 8x in a year.
According to Shopify data published via Yahoo Finance, in the first quarter of 2026 traffic from AI agents to stores on the platform grew 8x compared with the previous year. Orders generated by AI searches increased 13x. It's not an experiment: it's already the fastest-growing channel.
In Italy, according to the Netcomm NetRetail 2026 study, 67% of online shoppers use AI to compare products and prices, 66.8% to get personalized suggestions and 61.2% to summarize reviews. If an AI agent queries your store and finds an empty or wrong knowledge base, it simply leaves you out of the answer.
Since September 29, 2026, Shopify has enabled checkout through browser-based AI agents on all eligible stores: AI assistants can now complete purchases, including payment, without the user ever opening the site. The knowledge base is no longer a technical detail: it's your storefront for these new channels. You can learn more about AI agents in commerce in our article on personal AI agents and what changes for marketing in the coming years.
Where to start: the concrete steps to get your knowledge base in order
Six steps, in the right order. The first is understanding where you stand relative to the 1,000-fact limit.
- 1 · MerchantCheck your situationLog in to your Shopify admin and go to Sales channels > Agentic > Knowledge base. Check how many facts were in your previous app (also check the Shopify notification email).
- 2 · MerchantExport the facts from the app (if still accessible)If you have more than 1,000 facts, export or copy the app's content before it is shut down on January 15, 2027.
- 3 · Merchant / TeamReview and prioritize your factsNot all 1,000+ facts are equally useful. Select the critical ones: return policy, cancellations, shipping, warranties, payment methods.
- 4 · Merchant / TeamAdd the missing facts to the new adminUse the standard categories where possible. For information they don't cover (e.g. order cancellations), use the Additional Facts category.
- 5 · MerchantApprove the facts suggested by the scanReview the facts Shopify detects automatically and approve the correct ones. Reject or fix the inaccurate ones.
- 6 · Merchant / TeamTest with real questionsUse the Shop app or an AI tool to ask the questions your customers actually ask. Check that the answers are correct and complete.
What tools does Shopify use to let AI agents talk to your store?
Shopify uses two main technical tools: the Storefront MCP (now being phased out) and the new Ask MCP. Merchants don't need to know them in detail, but it helps to know what's changing.
MCP stands for Model Context Protocol. In plain terms, it's the channel through which an AI agent queries your store to get information. Shopify uses two versions.
| Tool | What it does | Current status | What it means for you |
|---|---|---|---|
| Storefront MCP (policy tool) | Provides policies and FAQs to AI agents | Deprecated (still works but will be removed) | Don't rely on it for new development |
| Ask MCP (ask_business tool) | Answers customer questions based on the knowledge base and policies | Active, in 'draft' (experimental) status | Availability varies across stores; it's the future |
| Admin knowledge base | Holds all the structured facts for AI agents | Active and mandatory from Jan 2027 | This is where you need to work now |
The fact that Ask MCP is in 'draft' status means availability may vary from store to store. That's no reason to wait: build your knowledge base now, so that when the tool is stable, your store is already ready. The topic of AI agents that operate autonomously, with both risks and opportunities, is explored in our article on Claude agent and autonomous AI, which is useful for understanding the boundaries of these systems.
What mistakes are most common in this migration?
The three most common mistakes are waiting until the last minute, publishing everything without checking, and forgetting use cases not covered by the standard categories.
| Error | Why it happens | How to avoid it |
|---|---|---|
| Waiting until January 14, 2027 | You think you have time | Start now: re-entering 1,000+ facts takes days of work |
| Publishing everything in bulk without reading it | The automatic scan seems reliable | Read every suggested fact: the scan can pick up outdated or wrong info |
| Forgetting order cancellations | There's no dedicated category | Add your cancellation policies as an Additional Fact |
| Not testing after the migration | You assume the transfer went fine | Test with real questions before considering the job done |
| Ignoring the 1,000-fact limit | You don't know the limit | Prioritize the most useful facts: quality > quantity |
| Leaving the storefront password-protected | Stores in development or staging | The scan won't work: enter the data manually or temporarily remove the protection |
How do I know if my AI knowledge base is working well?
There are direct and indirect signals. The new admin doesn't yet have a log of store questions, but you can use alternative methods to monitor the quality of the answers.
The old app had a metrics page with a log of customer queries to AI assistants. The new admin shows sales, orders, sessions and conversions by AI channel, but it doesn't record specific questions about the store. It's a temporary limitation that Shopify will probably fill, but in the meantime you need a different approach.
Tools like ChatGPT Work's Data Agent can help analyze customer service logs to identify the most frequent questions to add to the knowledge base as additional facts.
| Indicator | How to measure it | Warning sign |
|---|---|---|
| AI answer quality | Test manually with your customers' frequently asked questions | Vague or wrong answers, or 'I don't know' |
| Support escalation rate | Compare tickets before and after migration | More requests on topics the knowledge base covers |
| Conversions from the AI channel | Shopify Admin > Agentic Home | High AI sessions but low conversions → unsatisfactory answers |
| Facts in 'unpublished' status | Check the panel regularly | Facts suggested by the scan never approved → missing info |
| Uncategorized questions | Analysis of support tickets and chats | Recurring questions not covered → add them as Additional Facts |
Rules and risks to keep an eye on
The knowledge base is public by definition. Anything you publish can be read by any external AI platform. That's an advantage, but it calls for care.
Shopify states explicitly that the published knowledge base is publicly accessible through Storefront MCP. That means it's not just Shopify's Shop app: third-party AI assistants can query it too. It's an opportunity, since more AI channels can find you, but also a risk if the information is inaccurate.
Pay particular attention to: temporary promotional prices entered as facts (remove them once the promotion ends), policies being updated (don't publish interim versions), and information about products that are no longer available. The governance of data exposed to AI agents is covered more broadly in our article on Google and user data for AI training.
Another thing to keep in mind: the knowledge base doesn't affect how often your store appears in AI results. It improves the quality of answers, not visibility. To work on visibility in AI engines, you need a different path: SEO for AI systems, a topic we cover in the AI and process automation cluster.
This data, collected by Shopify in 2025, shows that the main problem isn't technology: it's a lack of direction. The knowledge base in the admin is a concrete, accessible starting point, even if you have no technical skills.
So, is it worth it for your company?
It's worth it if you have an active Shopify store with customers asking about policies, shipping or returns, and you want AI agents to answer correctly on your behalf. Even more so if you had more than 1,000 facts in the previous app: acting now prevents irreversible data loss.
Somewhere in between: you didn't use the app, but you have a storefront with complex policies or customized products. The automatic transfer covers the bare minimum, but enriching the knowledge base with additional facts concretely improves the quality of the AI answers your customers get.
Not yet if you're launching a store, with no significant AI traffic and simple policies already in the admin. In that case you can postpone advanced optimization, but keep an eye on the growth of the AI channel: the numbers move fast.
How we see it at SHM
For us, the AI knowledge base isn't a technical update to hand off to the IT department: it's the first content an AI agent reads before talking to your customers. We believe AI should be integrated to free companies from routine, and answering the same questions a thousand times is exactly that kind of routine. But for it to work, the information behind it must be correct, up to date and written with the reader in mind.
We've worked with Shopify e-commerce stores where the complexity of product variants and policies made it hard even to figure out what to tell human customers. Bringing that same clarity into an AI knowledge base takes the same information architecture work, except now the reader is a machine that won't forgive ambiguity.
If you run a Shopify store and want to know what to put in the knowledge base, how to structure it and how to test it, we can help with an audit and an action plan. We work on Shopify development and optimization, on the strategic integration of AI into sales processes and on the setup of virtual assistants that truly answer your customers' questions. You can also dig deeper into how AI agents are changing business data management in our article on Google Data Manager and AI agents.
Frequently Asked Questions
No, the automatic transfer has already happened and your facts have been published in the new admin. Still, you should check that everything is correct and complete, and approve any updates suggested by Shopify's automatic scan.
No. According to Shopify, the knowledge base improves the quality of the answers AI agents give about your products and policies, but it doesn't increase how often your store shows up in results. To work on visibility in AI systems, you need a dedicated SEO strategy.
If you had fewer than 1,000 facts, nothing critical happens: the transfer has already taken place. If you had more than 1,000 facts, those over the limit weren't transferred and will be permanently lost when the app is deactivated on January 15, 2027.
The new admin shows aggregated metrics by AI channel (sales, sessions, conversions) but doesn't record customers' specific questions about your store. For now, the best approach is to test manually with common questions and review your customer service tickets.
Yes. Shopify explicitly states that the published knowledge base is publicly accessible through the Storefront MCP, so any AI platform that supports this protocol can query it. That's a visibility advantage, but it means your information must always be accurate.
A password-protected storefront can't be scanned by Shopify. The knowledge base will only show the data you entered manually in the admin, without the facts automatically detected by the scan. You'll need to enter all the information by hand.
The AI knowledge base isn't internal documentation: it's the first thing an AI agent reads before talking to your customers.
Related articles
Discover more articles exploring similar topics, selected to offer you a more complete and stimulating perspective. Each piece of content is carefully chosen to enrich your experience.
