- AI assistants pick products from structured product data (feeds and catalogs) plus what the web says about you: reviews, comparisons, your own pages.
- The protocols for buying inside AI chats are still changing fast. Focus on what every assistant needs: accurate, complete, current product data and clear policies.
- Treat AI visibility like a new shopping channel. Fix your feeds first, then the content and third-party mentions that help assistants recommend you.
How are shoppers using AI assistants to buy?
More shoppers now start product research by asking an assistant instead of typing keywords: “best waterproof hiking boots for wide feet under $200” or “compare these three espresso machines”. The assistant returns a short list of products, often with images, prices and reasons.
For a store, that changes the question from “do I rank?” to “am I one of the three products the assistant suggests, and is what it says about me correct?”
This is part of a broader shift often called agentic commerce: AI agents that research, compare and in some cases complete purchases on behalf of shoppers.
Where do AI assistants get product information?
Assistants generally combine two kinds of sources.
| Source | Examples | What it provides |
|---|---|---|
| Structured product data | Merchant feeds, platform catalogs, product APIs | Titles, prices, availability, variants, images, shipping |
| Web content | Your product and category pages, reviews, comparison articles, forums, editorial coverage | Reasons to recommend, pros and cons, trust |
Structured data answers “what is available and at what price”. Web content answers “which one should I choose and why”. You need both. A perfect feed with no reviews or comparisons gives the assistant little reason to recommend you. Great reviews with a broken feed can lead to wrong prices or missing products.
What is the current state of buying inside AI chats?
This area is moving quickly, so treat the details below as a snapshot and check current status before making decisions.
- OpenAI launched Instant Checkout in ChatGPT in September 2025, together with the Agentic Commerce Protocol it developed with Stripe. In March 2026, OpenAI said it was deprioritizing Instant Checkout as a standalone product. ChatGPT shopping continues to show products and comparisons, and shoppers are sent to retailer sites to buy.
- Google announced the Universal Commerce Protocol (UCP) in January 2026 as an open standard for agent-led shopping, and in March 2026 added capabilities such as cart, catalog and identity linking. Google says it is bringing UCP capabilities to AI Mode in Search and the Gemini app, with onboarding through Merchant Center rolling out over time.
- Shopify introduced Agentic Storefronts in its Winter ’26 Edition, which let merchants share product data with AI platforms such as ChatGPT, Perplexity and Microsoft Copilot from inside Shopify.
The practical takeaway: whichever protocol wins, every assistant needs the same foundation. Accurate product data, clear policies and a strong reputation across the web.
What should you fix first?
1. Product feeds and catalogs
Your product feed is likely the main way assistants learn about your products, through Google Merchant Center, platform catalogs and integrations. Make sure:
- Titles describe the product clearly: type, key attributes, brand, model.
- Prices, availability and variants update frequently and match your site.
- Attributes such as material, size, color, gender, age group and compatibility are filled in.
- Images are clean and high quality.
- Shipping and return policies are set.
The work in Google Shopping feed optimization does double duty here.
2. Product pages assistants can read
Assistants and their crawlers read your pages. Make product and category pages easy to understand:
- Clear specs in text, not only in images.
- Accurate structured data for products, offers and reviews.
- Plain-language answers to common questions: sizing, compatibility, care, who it is for and who it is not for.
- Honest comparisons and “best for” statements, which help an assistant match your product to a specific request.
3. Access for AI crawlers
If your robots.txt blocks the crawlers that AI platforms use, your pages cannot inform their answers. Decide deliberately which crawlers to allow. See AI crawlers and robots.txt.
4. Reviews and third-party mentions
Assistants lean on what others say. Product reviews on your site, reviews on third-party platforms, mentions in comparison articles and discussions on forums all shape whether and how you are recommended. AI search citations for stores covers how to earn these mentions.
5. Clear policies
Shipping times, return windows and guarantees are things assistants mention when comparing stores. Put them in plain text on your site and in your feed settings.
How do you know if assistants recommend you?
Measurement is still immature, but you can build a practical routine:
- Prompt checks. Keep a list of 20 to 50 buying questions in your category (“best X for Y”) and run them in the main assistants monthly. Note whether you appear, what is said and whether prices and details are correct.
- Referral traffic. Watch referrals from AI assistants in GA4 and their conversion rate. Many assistants pass a referrer or UTM tags.
- Branded search trends. A rise in branded searches can reflect AI recommendations, since many shoppers search for a brand after an assistant mentions it.
- Post-purchase surveys. Add “AI assistant (ChatGPT, Gemini, etc.)” as an option to “How did you hear about us?”.
Dedicated AI visibility tools can automate prompt tracking. They are useful for trends, but results can vary from run to run, so look at patterns over time rather than single answers.
Worked example: an outdoor brand’s first AI visibility review
(Illustrative findings.) An outdoor gear brand runs 30 buying prompts across three assistants.
- It appears in 6 of 90 answers, mostly for its best-known jacket.
- Two answers quote an old price, and one says the jacket is not waterproof (it is).
- Competitors that appear more often have detailed spec tables, many reviews and mentions in several comparison articles.
Actions:
- Fix feed sync so price changes reach Merchant Center within hours.
- Add a spec table and a “waterproof rating explained” section to the jacket page.
- Run a review request campaign to past buyers.
- Pitch the jacket to two independent outdoor gear reviewers.
- Re-run the same prompts monthly and track appearances and accuracy.
Checklist: AI shopping readiness
- [ ] Product feed is complete, accurate and updated frequently
- [ ] Titles and attributes describe products the way shoppers ask for them
- [ ] Product pages show specs, sizing and compatibility in text
- [ ] Product, offer and review structured data validated
- [ ] AI crawler access decided deliberately in robots.txt
- [ ] Shipping, returns and guarantees stated clearly on site and in the feed
- [ ] Review collection running for key products
- [ ] Monthly prompt checks across main assistants with notes on accuracy
- [ ] AI referral traffic tracked in GA4
- [ ] “AI assistant” added as an option in post-purchase surveys
FAQ
Can I pay to appear in ChatGPT or Gemini shopping results?
Policies differ by platform and are changing. Organic recommendations generally depend on product data, relevance and reputation. Check each platform’s current merchant and advertising programs before assuming either way.
Is AI shopping visibility different from SEO?
It overlaps heavily. Good feeds, clear pages, structured data and a strong reputation help both. The main differences are the greater weight of feeds and catalogs, and the fact that assistants summarize many sources into one answer. See generative engine optimization.
Should I implement Agentic Commerce Protocol or UCP now?
For most small and mid-sized stores, the platform you use (Shopify, a feed tool or Merchant Center) will handle protocol integration. Focus on data quality and policies first, and enable integrations your platform offers once you understand what data they share and how orders are handled.