How to Get Recommended by AI Assistants
In short: no one can make an assistant recommend you. The work that holds up is making your site easy to find and read, publishing one clear page per question, getting described accurately by independent sources, and measuring the same questions over time. Each item below says how strong the evidence is.
This is the checklist we work through. After each item you will see where it comes from: official guidance (a company's own documentation), study (published research, usually correlational), our experience, or no evidence found.
Findable
- Allow the search crawlers you want to be found by. Your robots.txt and security settings should not block OAI-SearchBot (ChatGPT), PerplexityBot, Claude-SearchBot or Googlebot. OpenAI says a site that opts out of OAI-SearchBot will not appear in ChatGPT search answers. How to check each one. Official guidance.
- Get your key pages indexed. Google lists being indexed and eligible to appear with a snippet as a requirement for its AI features. Check Search Console, and Bing Webmaster Tools as good practice. Official guidance (Google); Bing is our habit.
- Reduce duplicate versions of key pages. Google recommends reducing duplicate content. Official guidance (Google). Where you can, put the main content in the page's HTML rather than only in scripts: Google says it can process JavaScript that is not blocked, but we have not verified that other assistants' crawlers run scripts. Our experience; no evidence found for other assistants.
- Skip llms.txt for now. Google says its Search does not use it. We have not found evidence that other assistants rely on it, so it should not come first. Official guidance; no evidence found for others.
Clear
- One page per question you want to be named for, such as a product, a use case or a comparison, with the question's wording in the title. In an Ahrefs analysis of about 1.4 million ChatGPT prompts, cited pages had titles closer in meaning to the prompt than pages that were not cited. Study, correlational, by a tool vendor.
- Answer in the first sentence, then give the detail: real specs, who it is for, who it is not for. Our experience; Google asks for people-first content with clear sections and headings.
- Use readable, natural-language URLs. Search results with such URLs were cited somewhat more often in the same Ahrefs data (about 90% against about 81%). It costs little, but treat it as a small signal. Study, correlational.
- Publish something only you can publish: your own tests, measurements, photos or first-hand experience. Google asks for original, first-hand, non-commodity content. Official guidance.
- Keep your name, description, specs and prices identical everywhere: your site, marketplaces, listings and social profiles. Our experience.
- Answer real follow-up questions in an FAQ. Use the questions customers actually send you. Our experience.
- Use structured data for the basic facts, but do not expect it to move recommendations. It still helps search engines read your organization and products. Ahrefs tracked 1,885 pages that added JSON-LD against matched controls for 30 days and found no clear citation gain, though every page was already heavily cited, so the test cannot say whether schema helps pages that are not yet cited. Google says structured data is not required for its AI features. Study by a tool vendor (observational, matched controls) plus official guidance.
- Do not mass-produce pages. Google calls large numbers of pages for query variations, made to manipulate rankings or AI responses, scaled content abuse. Official guidance.
Credible
- Be accurately described on the sites customers and assistants already trust: independent reviews, directories, marketplaces, community discussions, video. In the Ahrefs analysis of 75,000 established brands (Domain Rating above 40), mentions across the web and on YouTube tracked most closely with AI mentions. Study, correlational, by a tool vendor; it does not show that adding mentions causes visibility and says little about small or new brands.
- Know which comparison lists in your category matter, and be able to explain honestly why you belong on them. Our experience.
- Label your own comparisons as yours. A list you wrote about your own category is not independent, and the page should say so near the top. Our practice.
- Attribute claims you cannot prove ("according to us") and do not publish numbers you do not have. Our practice.
- Show who wrote a page, when it was updated and where its facts come from. Our experience.
Measured
- Keep a list of 20 to 50 buyer questions, written without your brand name. Our practice (see how we test).
- Report questions that name your brand separately. A brand usually looks better when you ask about it by name than when you ask a buyer question. Our practice.
- Run the same questions on a regular schedule across several assistants, and track how often you are named rather than your exact position. SparkToro and Gumshoe found that the odds of getting the same list of brands twice were under 1 in 100. Study, run with a company that sells AI-tracking tools.
- For Google's AI features, use the Generative AI performance report in Search Console. Google also warns that no third-party tool has access to its internal ranking or AI systems, so check tool advice against Google's guidance. Official guidance.
- After any change, compare like with like before deciding it worked. Our practice.
Honest
Fake reviews, hidden text, paid mentions presented as independent and promises of guaranteed placement may look like shortcuts, but they put your credibility at risk. Google's guidance says seeking inauthentic "mentions" across the web "isn't as helpful as it might seem". The work that lasts is the plain version: clear pages, accurate facts, real proof and steady measurement.
If you only have a week
- Pick three specific questions.
- Run them on four assistants, three times each, and save the results as your baseline before you change anything. Also ask each assistant "What is [your brand] and what do they make?" and note any wrong facts.
- Confirm the search crawlers are allowed and your key pages are indexed.
- Write one clear page for each question, and fix any wrong facts you found when you asked about your brand by name.
- Run the same questions again in a few weeks and compare like with like.
Questions people ask
How long does it take to get recommended?
Nobody can give a reliable timeline, and we do not promise one. Set a baseline, make changes, then rerun the same questions after a few weeks and compare.
Should I block GPTBot or ClaudeBot?
Those crawlers relate to model training, not search, so blocking them is a separate business decision. OpenAI names OAI-SearchBot as the crawler that controls whether a site appears in ChatGPT search answers, and Anthropic names Claude-SearchBot for search.
Does schema markup help AI assistants recommend me?
It helps machines read your basic facts, but we have not seen evidence that it changes recommendations. One Ahrefs test of 1,885 pages found no clear citation gain after adding JSON-LD.
Can I just ask an agency to do all of this?
You can, but ask what they measure, how many runs they use and whether they promise placement. What Are AEO and GEO? has a list of questions to ask a provider. How to Choose an AI Visibility Agency goes through ten questions and a scorecard.
Sources
- OpenAI, overview of OpenAI crawlers: OAI-SearchBot, GPTBot, ChatGPT-User.
- Perplexity, crawler guide: PerplexityBot and Perplexity-User.
- Anthropic, how its crawlers work.
- Google, optimizing for generative AI features in Search, updated Jul 10, 2026.
- Ahrefs, why ChatGPT cites the pages it does, Apr 15, 2026: about 1.4 million prompts one dataset of desktop prompts (Ahrefs gives the month as February 2025 but also names ChatGPT 5.2, so the month looks like a typo); vendor study, correlational.
- Ahrefs, AI brand visibility correlations, Dec 12, 2025: 75,000 established brands (Domain Rating above 40); vendor study, correlational.
- Ahrefs schema markup test, reported by Search Engine Journal, May 11, 2026: vendor study, 1,885 pages, 30 days before and after, matched control pages.
- SparkToro and Gumshoe study on repeat AI recommendation lists, Search Engine Land, Jan 28, 2026. Run with Gumshoe.ai, an AI-tracking startup that sells such tools, and described by Search Engine Journal as not peer-reviewed research. Volunteers ran the prompts in their own AI settings. Tested ChatGPT, Claude and Google's AI.
Related guides
We show you which pages AI reads in your field, then write the comparison page you're missing, ready to publish.
Get your free comparison page