How to choose a UGC platform when AI assistants do the shopping

On 1 October 2026, OpenAI added virtual try-on and Favorites to ChatGPT shopping results. TechCrunch reported it the same day. In March, OpenAI stepped back from Instant Checkout. Merchants now use their own checkout, and can share product feeds with OpenAI.

So the assistant is where a shopper finds a product and pictures it. Your site still takes the order. We covered the launch in ChatGPT can try on your products now.

Review and UGC vendors moved in the third quarter. Bazaarvoice launched an AI Visibility package on 1 September. Videowise launched AI Visibility on 3 September. Whatmore added AI Discovery in its Q3 2026 edition. Okendo and Tolstoy now run MCP servers. The questions below help you compare the pitches.

How to use this guide

Each section is one question to put to a vendor. The checklist table shows a good answer and a red flag for each. Vendor facts come from their own pages and press releases, read 13 October 2026. We make Idukki, and we answer the same questions near the end.

1. Are my reviews in the page HTML as structured data?

Bazaarvoice's launch release says most review widgets render in JavaScript. It says many AI crawlers skip that format. A crawler that skips the script never sees your reviews.

Ask for Product, Review and AggregateRating JSON-LD in the raw HTML. Then check it yourself. Open the page source and search for "aggregateRating". Our review schema validator runs the same check.

Ask about Google too. A product reviews feed in Merchant Center format sends your ratings to Shopping. Ask whether the vendor builds one, and from which reviews.

2. Can an assistant fetch my content without running a script?

Structured data helps when a crawler loads your page. A feed helps when an agent asks for data directly. Ask whether the vendor publishes a feed for each store. It should carry products, reviews and media together, as JSON or JSON-LD.

Ask about llms.txt as well. It is a short text file that points AI crawlers to your key pages. It is a proposal at llmstxt.org, so treat it as cheap insurance. Our llms.txt guide covers the file.

3. Does it offer MCP access, and for whom?

MCP is the Model Context Protocol. It lets an AI assistant call tools in another app. Two kinds of MCP server matter here. Vendors often blur them.

• A merchant MCP server. You connect it to your own Claude or ChatGPT. It reads your reviews, posts or sales.
• A shopper-agent endpoint. A public, read-only feed for AI shopping agents. It should serve only content you have cleared.

For a merchant server, ask four things. What can it read? Is write access off by default? How do you revoke a connection? Does every write land in an audit log?

The answers vary a lot. Okendo's server needs no sign-in, because it reads only published reviews and Q&A. Tolstoy signs in with OAuth 2.1, and its agent can publish media. Those are different risks. Our MCP comparison puts four servers side by side.

4. Is the media cleared for reuse?

An assistant that quotes a review or shows a customer photo is reusing it. A public post is not a licence. Ask how the vendor records consent for each post. Ask whether its feed or agent endpoint can serve posts without consent.

Then ask what happens when a creator withdraws. The answer you want: the post leaves every gallery and feed at once. What is UGC rights management covers the process.

5. Is each photo tagged to the right product?

A customer photo tied to a SKU is evidence an assistant can attach to that product. An untagged photo is just a picture. Ask how posts get tagged: by hand, by AI or both.

If AI tags them, ask what it does when unsure. The answer you want is a confidence threshold with a review queue. Ask about variants too. A photo of the green jacket should not sit on the blue one.

6. Does it work on every store you run?

Several of the new AI products name one platform. Videowise's AI Visibility tour says it is for Shopify only. Whatmore's AI Discovery ships in its Shopify app. Running WooCommerce, Magento or Salesforce Commerce Cloud as well? Ask before you sign.

7. What does it add to the product page?

An assistant sends the shopper to your page. That page still has to load on a phone. Every gallery and review widget adds a script. Ask for the gzipped size of the core download, measured on a live page.

Idukki's core download is 79 KB gzipped, measured in October 2026. Test any vendor in PageSpeed Insights with the widget on and off.

8. Can it show whether assistants cite you?

Ask this one last. Citation tracking means running real prompts in ChatGPT, Gemini or Perplexity. Then you count where your brand appears. Yotpo Discover does this. Its initial release is for brands with $10M+ in annual GMV. Bazaarvoice offers tracking through partners: Bluefish, Evertune and Profound.

Most UGC platforms, ours included, do not run prompts. Ask any vendor that does which engines it checks, with how many prompts and how often. Fix what crawlers can read before you pay to count. Measuring first only measures the gap.

The checklist in one table

Where Idukki stands (ours)

We make Idukki, so here are our own answers. On Shopify, product pages get Product, Review and AggregateRating JSON-LD. An opt-in Google product reviews feed sends approved, rated reviews to Merchant Center. It stays off until you switch it on.

Each store gets a public agent feed. It serves products with rights-cleared UGC, reviews and the aggregate rating. It comes as JSON, as JSON-LD and as an llms.txt file. A read-only MCP endpoint serves the same data to shopping agents. Our agentic commerce page lists that public side as beta.

A private MCP server connects your own assistant. It has 12 read tools and 3 write tools. New keys are read-only. On claude.ai and ChatGPT, write access is an unticked box. Every write attempt goes in the audit log. You can revoke a key or app in Settings.

Idukki runs on Shopify, WooCommerce, BigCommerce, Magento, PrestaShop and SFCC. The gap: Idukki does not run prompts or track citations. If you need that number, pair it with a tracker.

Comparing named vendors? Read the best AI visibility tools for reviews and UGC. Choosing the gallery itself? See how to choose a UGC gallery app.