What people mean by AI search visibility

"AI search visibility" is not a term any search engine uses in its own documentation. In practice it covers two questions: whether your pages are used and linked when an AI system answers a question, and whether your products appear when someone shops through one of those systems. For a Shopify store, the systems worth understanding fall into three groups.

  • Google's AI features in Search. AI Overviews appear above the ordinary results for some queries; AI Mode is a separate, conversational search experience for questions that need more exploration or comparison. Both link to the pages they drew on.
  • AI assistants that search the web. ChatGPT search, Microsoft Copilot and Perplexity fetch and cite pages when they answer, each with its own documented crawlers and controls.
  • Shopping and agentic channels. These take product data directly rather than reading your pages. Google Merchant Center feeds Google's shopping experiences, and Shopify now sends product data to several AI channels through Shopify Catalog.

The distinction matters because the controls differ. A robots.txt rule tells crawlers what they may fetch. It does nothing to a product feed you have chosen to send.

How Google's AI features choose pages

Google's page on AI features and your website is unusually direct. It says: "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary." To be eligible as a supporting link, a page needs to be indexed and eligible to be shown in Google Search with a snippet. That is the whole entry requirement, and Google adds that meeting it does not guarantee a page will be crawled, indexed or shown.

The same page rules out several things that are widely sold as AI optimisation. In Google's words: "You don't need to create new machine readable files, AI text files, or markup to appear in these features," and "There's also no special schema.org structured data that you need to add."

What Google recommends instead is the SEO fundamentals: crawling allowed in robots.txt and by any CDN or hosting layer, internal links, a good page experience, important content available as text, structured data that matches the visible text, and up-to-date Merchant Center and Business Profile information. That is a technical SEO checklist, and that is the point.

Google also says both features may use "query fan-out", running several related searches across subtopics while building a response, so they can link to a wider range of pages than classic search. Our reasoning, not Google's statement, is that a product page answering specific questions (is it machine washable, does it ship to Ireland) gives those searches more to match. Traffic from both features is reported in Search Console's Performance report under the "Web" search type, inside the figures you already have.

The crawlers behind AI assistants

Each AI company publishes the names of its crawlers, and most now separate crawling for search from crawling for model training. These are the ones a store owner is most likely to meet, described as their owners describe them.

Crawler or controlWhoseWhat it governs, according to the ownerDocumented at
GooglebotGoogleCrawling for Google Search, which AI Overviews and AI Mode draw onGoogle crawler list
Google-Extended (robots.txt token, not a separate crawler)GoogleUse of your content to train Gemini models and for grounding in Gemini Apps and Vertex AI. Google says it does not affect Google SearchGoogle crawler list
OAI-SearchBotOpenAIWhether your site can be shown in ChatGPT search answersOpenAI crawlers
GPTBotOpenAIUse of your content to train OpenAI's modelsOpenAI crawlers
ChatGPT-UserOpenAIFetches a ChatGPT user asked for. OpenAI says robots.txt rules may not applyOpenAI crawlers
PerplexityBotPerplexitySurfacing and linking sites in Perplexity's results. Perplexity says it is not used for model trainingPerplexity crawlers
Perplexity-UserPerplexityFetches a user asked for. Perplexity says it generally ignores robots.txtPerplexity crawlers
NOARCHIVE and NOCACHE meta tagsMicrosoftHow much of a page Bing Chat (now Copilot) and Microsoft's model training may use. Pages stay in Bing search either wayBing Webmaster Blog, September 2023

The pattern that matters is the split between search and training. OpenAI says of its crawlers: "Each setting is independent of the others." Google says Google-Extended is not a ranking signal and does not affect inclusion in Google Search. So a store can keep its content out of model training and remain eligible for ChatGPT search and Google's AI features.

Two cautions. Blocking Google-Extended does not keep a page out of AI Overviews; for that, Google points to the Search snippet controls covered below. And the user-triggered fetchers are documented as not reliably following robots.txt, because a person asked for the page.

Microsoft announced its NOARCHIVE and NOCACHE controls in September 2023, before Bing Chat became Copilot. In February 2026 it added an AI Performance report to Bing Webmaster Tools, which shows how often your pages are cited in Copilot, in AI-generated summaries in Bing and in some partner integrations. If your store is not verified in Bing Webmaster Tools, that is a good reason to do it.

What a Shopify store controls

robots.txt, through robots.txt.liquid

Shopify generates a robots.txt for every store. To change it you add a robots.txt.liquid template to the theme, which replaces the generated file. Shopify's help centre calls this an "unsupported customization", says Shopify Support cannot help with it, and warns that "Incorrect use of the feature can result in loss of all traffic." Shopify's developer documentation recommends keeping the Liquid that outputs the default rules, adding your own groups after it, and comparing the result with a saved copy of your current /robots.txt.

If you decide to keep training crawlers out while leaving search crawlers alone, the template looks like this:

{% for group in robots.default_groups %}
  {{- group.user_agent }}

  {%- for rule in group.rules -%}
    {{ rule }}
  {%- endfor -%}

  {%- if group.sitemap != blank -%}
    {{ group.sitemap }}
  {%- endif -%}
{% endfor %}

User-agent: GPTBot
Disallow: /

User-agent: Google-Extended
Disallow: /

This shows the mechanics, not a recommendation. Whether your product copy should be used for model training is a business decision, and none of these companies documents a link between allowing training crawlers and appearing in their search answers. OpenAI and Perplexity both say robots.txt changes can take around 24 hours to take effect.

Shopify adds that bot management at the network level is handled for you, and that robots.txt rules are "directional and advisory": well-behaved crawlers follow them, but nothing forces the rest to.

Snippet controls

Google's robots meta tag documentation is now explicit that its snippet controls reach AI features. nosnippet applies to all forms of search results, including AI Overviews and AI Mode, and "will also prevent the content from being used as a direct input" for them. max-snippet limits how much of a page can be used in the same way. The data-nosnippet attribute excludes a specific part of a page, and works only on span, div and section elements.

For a shop, nosnippet on a product page is a blunt instrument, because it also removes your ordinary snippet in Google. data-nosnippet is more useful for a block you would rather not see quoted out of context, such as promotion small print:

<div data-nosnippet>
  Offer ends Sunday. Excludes sale items and gift cards.
</div>

Changes apply once the page is recrawled; URL Inspection in Search Console shows when that has happened.

Product data: structured data, Merchant Center and Shopify Catalog

Ecommerce has an advantage here that content sites do not: structured product data that machines already know how to read.

Most Shopify themes output Product structured data. Google's product structured data documentation separates product snippets, for pages where people cannot buy directly, from merchant listings, for pages where they can, and says that providing both structured data and a Merchant Center feed "maximizes your eligibility to experiences and helps Google correctly understand and verify your data". It is equally clear that enhanced results are shown at Google's discretion and are never guaranteed.

The Merchant Center product data specification is strict about consistency. Price and availability in the feed should match the landing page, the structured data and the checkout. GTINs should be included where they exist and never guessed. Products with no brand should leave the brand empty rather than say "Generic". On Shopify, the Google & YouTube sales channel is the usual way to send products to Merchant Center, and mismatches show up there as product issues or disapprovals.

Shopify's agentic storefronts are the newer piece: customers discover and buy your products in AI channels, which Shopify names as ChatGPT, Google AI Mode and Gemini, Microsoft Copilot and Meta. Products reach them through Shopify Catalog, built from the product data already in your admin. Four points from Shopify's documentation:

  • It is active by default for eligible stores. Shopify does not set out eligibility on the pages we read, so check under Sales channels, Agentic in your own admin to see whether your store is included and which channels are on.
  • Blocking AI crawlers in robots.txt does not switch it off. Shopify says that blocking at the robots.txt or network layer "affects only open-web discoverability", and that Catalog sharing is managed in the agentic storefront settings.
  • A knowledge base in those settings holds non-product information such as returns and shipping.
  • Shopify's agentic settings include a "Shop policy completeness" signal that checks whether your shipping, returns, refund and other policies are present, and Shopify says: "Policies signal store legitimacy and can help AI channels confidently recommend your products." Shopify also says inclusion in Catalog does not guarantee a product will appear in any particular AI answer.

What is not known, and what to be sceptical of

Nobody can guarantee that an AI system will cite your store. Google and Shopify both say as much in their own documentation, and Microsoft says its "average cited pages" figure reflects overall citation patterns, not a page's ranking or authority. A provider who promises a number of AI citations is promising something none of these companies offers.

There is no special "AI SEO" markup that Google requires, as quoted above. Structured data helps machines understand a product page, and Google recommends that it match the visible text, but it is not a separate ticket into AI answers.

llms.txt

The llms.txt idea is a proposal published by Jeremy Howard in September 2024: a Markdown file at /llms.txt with the site's name, a short summary and lists of links to the most useful pages, plus clean Markdown versions of those pages. The site describes it as a proposal open to community input. It reports adoption by thousands of sites and documentation tools, but it does not claim that any search engine or AI assistant uses the file to decide what to show or cite. Google's AI features page says you do not need "AI text files", and none of the crawler documentation listed in our sources mentions reading one.

If you decide to add one, treat it as low-cost housekeeping rather than a visibility tactic, and check first whether your store already serves a file at /llms.txt.

Measurement

Tools that sample AI answers and report a "visibility score" are reporting on the prompts they chose to run: useful for tracking change, provided you know which prompts. First-party data is thinner: Google counts AI feature traffic inside Search Console's web search figures, Bing Webmaster Tools reports citations, and visits from assistants reach your analytics only when someone clicks through and the referrer is passed, so treat those figures as a minimum.

A sensible preparation plan for a Shopify store

Almost everything worth doing already pays off in ordinary search. This is the order we would follow.

  1. Confirm the store can be crawled and indexed

    Read your live robots.txt, check key templates for stray noindex tags, and look at the Pages report in Search Console. Verify the store in Bing Webmaster Tools too.

  2. Decide your crawler policy on purpose

    Choose whether search crawlers and training crawlers may fetch the store, write the decision down, and only then edit robots.txt.liquid if anything needs to change.

  3. Put the answers on the page as text

    Sizes, dimensions, materials, compatibility, care, delivery times and costs by destination, and the returns window. Specific facts in text, not only in images or tabs that load late.

  4. Make the data agree everywhere

    Price, availability, GTIN and brand should match across the product page, its structured data, the Merchant Center feed and checkout.

  5. Review your agentic storefront settings

    See which AI channels are active, whether direct checkout is on, and fill in the knowledge base and store policies.

  6. Measure with first-party data

    Use Search Console, the Bing AI Performance report and analytics referrals, and record a dated baseline before changing anything.

A few notes on the steps that cause most trouble in practice.

The crawl check catches more than you would expect: staging rules left in a custom robots.txt.liquid, a theme adding noindex to collections, or a security app blocking unfamiliar user agents. If you are changing theme or platform, our migration SEO checklist covers the launch checks.

Product information is where most stores have the most room to improve. Many product pages repeat the manufacturer's description and keep the useful facts in a size chart image or a PDF. A system answering "which of these is waterproof and fits a 15 inch laptop" can only use what it can read. The same goes for policies: a delivery page that states costs and times for each country you ship to answers questions a vague one cannot.

Consistency is the dull step that matters. If the page says in stock, the structured data says out of stock and the feed has last week's price, every system reading that product has to guess which is right.

Rewriting product descriptions at scale is a workflow question as much as a writing one; our AI workflows work covers drafting with human review so that nothing reaches a product page unchecked. If you would like a second opinion on where your store stands, get in touch.

Sources