AI Search Engines Compared: How Each Finds Your Store

For a store, the useful difference between AI search engines is access, not answer quality. Each one has its own crawler, index and rules for which pages it may quote. We read the documentation of seven AI search surfaces on 6 October 2026 and put the parts a store controls side by side, with the order to fix them in.

Why compare surfaces instead of brands?

Because one company can run several AI products that follow different rules. Google alone runs AI Overviews, AI Mode and the Gemini app, and they are not gated by the same robots.txt token. Comparing "Google" with "ChatGPT" hides the settings that decide whether your store can be quoted at all.

Most of the pages ranking for this query compare engines as tools to use: which one writes the best summary, which subscription is cheapest, which has the largest context window. That answers a shopper's question. A brand needs the reverse view: what each engine reads, and what stops it from reading you.

We hold ourselves to the same discipline when we measure. Our GEO measurement protocol requires the product mode to be named exactly, because "ChatGPT Search is not the ChatGPT API, and AI Mode is not AI Overviews". The same protocol lists "one tactic works across all engines" as a claim with no evidence behind it, noting that ChatGPT, Perplexity, Google, Claude and Copilot "have documented differences in crawlers, retrieval, and product behaviour."

The surfaces side by side

The table covers what a store controls or can observe on each surface. AI Overviews and AI Mode share every control, so seven surfaces fit in six rows. Cells come from each platform's own documentation, read on 6 October 2026, with two exceptions: the Claude and Brave link (a 2025 report) and the Shopify routes (Shopify's changelog).

The Google, Claude and Copilot rows are the ones stores most often get wrong, so they come next.

Google runs three AI surfaces with two different switches

AI Overviews and AI Mode are part of Google Search, so ordinary Googlebot access decides whether you can appear. The Gemini app is different. A separate robots.txt token, Google-Extended, decides whether Gemini may use your pages to ground its answers, and blocking it leaves your AI Overviews eligibility untouched.

For AI Overviews and AI Mode, Google's AI features guide asks for nothing beyond a page that is "indexed and eligible to be shown in Google Search with a snippet". Its one merchant-specific instruction is to keep your Merchant Center and Business Profile information current.

The Gemini app is where stores trip. Google's crawler documentation gives Google-Extended two jobs: training future Gemini models, and grounding in the Gemini app, meaning "providing content from the Google Search index to the model at prompt time". A store that blocked Google-Extended to keep its copy out of training has also opted out of being live evidence in Gemini app answers. If you made that block, check you meant to give up those citations.

We cover the Shopify robots.txt mechanics in our guide to GEO for ecommerce; the short version is that the training and search tokens are separate at OpenAI and Anthropic too.

Claude has used Brave, and Brave follows Googlebot

Claude's search has two documented routes to your pages: Anthropic's own Claude-SearchBot, and Brave Search, which Anthropic listed as a subprocessor in March 2025. Brave will not crawl a page Googlebot cannot. So a Googlebot block leaves Claude-SearchBot free to read the page but closes the Brave route to it.

Anthropic documents three agents. ClaudeBot collects training data, Claude-User fetches pages when a user asks, and Claude-SearchBot indexes content for search. Anthropic warns that disabling Claude-SearchBot "may reduce your site's visibility and accuracy in user search results."

The Brave link is the older evidence. It comes from Simon Willison's report of Anthropic's subprocessor list in March 2025; Anthropic's trust center would not load for us on 6 October 2026, so we could not confirm the entry is still there. Brave's side is current: its crawler page says "if a domain or page is not crawlable by Googlebot, then Brave Search's bot will not crawl it either," and Brave has no separate user agent to allow. If Googlebot is blocked on a section of your store, assume Claude can reach it only through its own crawler.

Only Bing shows you your AI citations

Of the surfaces in the table, Microsoft is the only vendor we found on 6 October 2026 that gives site owners a free report of AI citations. The AI Performance report in Bing Webmaster Tools, in public preview since February 2026, lists cited pages and the grounding queries behind them.

Microsoft's launch post says the report covers "Microsoft Copilot, AI-generated summaries in Bing, and select partner integrations", with total citations, cited pages per day, page-level counts and the key phrases the AI used when retrieving your content. Read it as directional: Microsoft says "the data shown represents a sample of overall citation activity."

Google gives you less. Its AI features guide says traffic from AI Overviews and AI Mode is counted within the "Web" search type of Search Console's Performance report, so you see the clicks but cannot separate them from classic results. Our GA4 setup for AI traffic catches visits from chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and claude.ai, and it can never catch AI Overviews, whose clicks arrive as plain google.com referrals.

For ChatGPT, Perplexity and Claude we found no first-party citation report. There, the only way to know whether you are cited is to ask the engine your buyers' questions and record what comes back, repeatedly, which is what our measurement protocol is built for.

Where Reddit and other forums count

A Reddit or community plan pays off on four surfaces and barely registers on two. Reviewing third-party citation analyses, our claims audit concluded that Gemini, AI Mode, AI Overviews and Perplexity cite Reddit heavily, while ChatGPT and Claude effectively do not. One blended "AI cites Reddit" figure hides that split.

ChatGPT is the instructive case, because it does see Reddit. The analysis summarised in our GEO claims audit found Reddit pages appearing as candidate sources in 76% of ChatGPT searches, but only 0.61% of those pages were cited. If your buyers ask Perplexity and Google, genuine forum presence can surface. If they ask ChatGPT, editorial reviews, comparison articles and your own product pages are the sources in play.

This split moves, too. The same audit lists "AI citations are stable once earned" as false, citing a 2025 Semrush study in which ChatGPT's share of answers citing Reddit fell steeply over six weeks. Our four-versus-two reading is from August 2026; check it on your own prompts before you plan a quarter around it.

How often the engines search at all

Whether an engine searches the web depends heavily on intent, and the rates move. In one vendor's ChatGPT study, commercial prompts triggered a web search almost every time and informational prompts almost never. Buying questions are the ones most likely to pull in live pages, which is where a store's product data can be cited.

The study by cloro, a company that sells AI visibility tracking, reports that "across 634 responses in August 2026, ChatGPT ran a web search on 86.5% of commercial prompts and 0.9% of informational prompts." It is one vendor's measurement on one prompt set, so treat the size as indicative. "Best merino hoodie under €100 that ships to Germany" needs current prices and stock; "what is merino wool" does not.

The same company's comparison of six engines, run on roughly 2,500 prompts each at three readings, found that ChatGPT "fell from top to second-bottom between July and August" and had recovered by October, so no engine's rate is fixed. The one result that held: "Gemini grounds least of the six on all three dates." On Gemini, fewer answers are built from a live page, so a fresh product page has fewer chances to be read there than on the other five.

The order to work through them for a DTC store

Start with access, because every other effort is wasted on a surface that cannot read you. Then Google, where your existing index and Merchant Center data carry over. Then ChatGPT, then Bing for the free citation data, then the forum and review sources that matter only on some engines.

Take an example store selling a €79 merino hoodie that ships to the EU and the US, and a buyer asking "best merino hoodie under €100 that ships to Germany". Here is how we would work the surfaces for that page:

1. Run the access checks. Open your robots.txt and confirm nothing disallows OAI-SearchBot, PerplexityBot, Claude-SearchBot, Googlebot or Bingbot, and that a Google-Extended block, if you have one, is a choice you made on purpose. Allow the user-triggered agents too: OpenAI says ChatGPT "may visit a web page with a ChatGPT-User agent", and Anthropic warns that disabling Claude-User "may reduce your site's visibility for user-directed web search." Then check your CDN or firewall: Perplexity's crawler documentation asks you to permit PerplexityBot's published IP ranges, and says Perplexity-User "generally ignores robots.txt rules", so a firewall is the only place it can be blocked by accident. The most common technical error we find in audits is a site allowing GPTBot and assuming that makes it eligible for ChatGPT Search; OpenAI's crawler overview is explicit that OAI-SearchBot is the token that controls ChatGPT search answers.
2. Make the hoodie page work for Google. The page must be indexed and allowed a snippet. Its €79 price and the EU and US shipping regions should be in the page text so an answer can quote them, and the Merchant Center feed should match. That covers AI Overviews and AI Mode, and the Gemini app if Google-Extended is open.
3. Get the hoodie into ChatGPT and Copilot. Shopify says products are "automatically accessible to AI channels like ChatGPT, Shop, and Copilot through Shopify Catalog" in its developer changelog; other stores can apply for access to OpenAI's product feed through OpenAI's commerce onboarding page. Our guide to the ChatGPT product feed compares the routes.
4. Verify the store in Bing Webmaster Tools. Then watch whether the hoodie page shows up among cited pages in AI Performance, and for which grounding queries.
5. Place off-site effort by engine. If Perplexity and Google matter for merino hoodies, earn genuine mentions in outdoor and travel-gear forum threads and review sites. For ChatGPT and Claude, put the same effort into editorial roundups and comparison articles that review hoodies in your price band.
6. Measure per surface, not in one blended number. Ask "best merino hoodie under €100 that ships to Germany" and a few sibling questions on each surface in fresh, logged-out sessions, and keep each surface's results apart. Our guide to measuring AI visibility covers how many runs you need before a change means anything.

Steps 1 and 4 cost nothing. Do them first, then let the measurement in step 6 tell you which of the remaining surfaces deserves your budget.