How to get your app recommended by ChatGPT

The 7 steps

1. Let the AI search crawlers in

   ChatGPT and Perplexity use their own search crawlers, and blocking them keeps you out. OpenAI says sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers, and recommends allowing it in robots.txt. Perplexity recommends the same for PerplexityBot. For Google's AI Overviews and AI Mode, a page must be indexed and eligible to show in Google Search with a snippet.

   • OAI-SearchBot (search) and GPTBot (model training) are separate settings. You can allow search and still block training.
   • OpenAI says robots.txt changes take about 24 hours to reach ChatGPT search. Perplexity says up to 24 hours.
   • Google says preview controls like nosnippet and max-snippet can limit how your content shows in AI formats, so check them on key pages.

   User-agent: OAI-SearchBot
   Allow: /
   
   User-agent: PerplexityBot
   Allow: /

2. Get onto the pages engines cite

   ChatGPT search, per OpenAI, rewrites a question into one or more targeted queries and sends them to third-party search providers. Google's AI features ground answers in its Search index and run several related searches at once. Perplexity searches the web and cites numbered sources. So the pages that rank for best-app searches in your category are the ones the answer is likely built from.

   • Ask the engines your buyer questions and write down which domains they cite. Those are your targets.
   • Sort them by type (comparison articles, review sites, app directories, forums). Pitch the editors, earn genuine reviews, and correct wrong facts about you.
   • Skip paid fake mentions. Google tells site owners to ignore inauthentic mentions.

3. State your category plainly on your own site

   An engine can only repeat what it can read. Put one plain sentence high on your homepage that says what the app is, who it is for and what it does differently, and use the same wording on your About page and your store listings. Keep the name, category and core facts identical everywhere so every source agrees.

   • A sentence like "[App] is a [category] app for [who] that [does what]." is easy for a model to lift.
   • Apple ranks store search partly on your title, subtitle and primary category, so the same sentence helps there too.

4. Answer buyer questions on your site

   Write the questions buyers actually ask (is it free, does it work offline, how is it different from the big name) and answer each one directly. Google describes query fan-out, where its AI runs several related sub-queries to answer one question. A page that answers those sub-questions plainly gives the engine something to use.

   • Lead each answer with the direct answer, then the detail.
   • Google says structured data should match the visible content, and that you do not need special schema to appear in its AI features. Write for people first.

5. Use the tactics the research measured

   The original GEO paper (Aggarwal et al., arXiv 2311.09735, KDD 2024) rewrote source pages and measured how often generative engines used them. Citing sources, adding quotations and adding statistics did best, with a 30 to 40% relative gain on its main visibility measure. Clearer, more fluent text helped by 15 to 30%. Keyword stuffing gave little to no gain, and on Perplexity it did 10% worse than the unchanged page.

   • Add a real number where you would write an adjective ("rated 4.8 by 12,000 users" beats "loved by users", if it is true).
   • Quote a credible source or a real customer, with a name.
   • Link to where your claims come from.

6. If you ship a ChatGPT app, tune its metadata

   Apps inside ChatGPT, which OpenAI's docs now call plugins, get picked a different way. OpenAI says ChatGPT decides when to call your tool based on the metadata you provide, and that good names, descriptions and parameter docs increase recall on relevant prompts. OpenAI has not published how its plugin directory ranks apps, and listing requires passing its review.

   • Start each tool description with "Use this when..." and say what it should not be used for, as OpenAI recommends.
   • Build a golden prompt set of direct, indirect and negative prompts, and test against it.
   • Change one metadata field at a time and review tool-call analytics weekly.

7. Measure across engines, several times

   AI answers change from one run to the next, so one check tells you almost nothing. Write down the questions your buyers ask, run each one several times on each engine, and count how often your app gets named, where it appears in the answer, and which sites get cited. Repeat on a fixed schedule so each run compares to the last.

   • Use the same prompts every time, typed the way buyers type them.
   • Track ChatGPT, Perplexity, Gemini and Claude separately. They can disagree.
   • For Google, check the Generative AI performance report in Search Console.

How to measure whether ChatGPT recommends your app

Measurement tells you whether the other six steps worked. The core idea is to sample, because a single answer is an anecdote.

This is how our own tracker at Agniverse GEO runs. Each answer is recorded with the brands it named, in what order, and the sites it cited. From that you get three numbers worth watching: how often you are named (your mention rate), where you sit in the answer, and which domains the engines lean on for your category.

You do not need our setup to do this. A spreadsheet, ten buyer questions and three runs per engine is a fair start. The paid AI visibility tools automate the same loop, and our cost guide lists their published prices.

Before you call it done

A quick pass to confirm the basics are in place:

• robots.txt allows OAI-SearchBot and PerplexityBot.
• Your key pages are indexed in Google and not blocked with nosnippet.
• One plain sentence says what the app is, the same on your site and store listings.
• A page answers the questions buyers ask, each with a direct first line.
• Claims on key pages carry a source, a quote or a number.
• You have a list of the domains the engines cite for your category, and a plan for each.
• If you ship a ChatGPT app, its tool descriptions start with "Use this when...".
• A fixed prompt set exists, with a baseline from several runs per engine.

Frequently asked questions

Sources

The primary source behind each step.

1. OpenAI, Overview of OpenAI crawlersOAI-SearchBot vs GPTBot; opt-outs; about 24 hours for robots.txt changes.
2. OpenAI Help Center, ChatGPT searchRewrites questions into queries for third-party search providers; citations can be incomplete.
3. OpenAI, Optimize metadata (plugins)ChatGPT decides when to call a tool from its metadata; golden prompt sets.
4. OpenAI, plugin submission guideReview and publishing flow for the ChatGPT plugin directory.
5. Perplexity, Perplexity crawlersPerplexityBot surfaces sites in Perplexity; changes within 24 hours.
6. Perplexity Help Center, How does Perplexity workSearches the web; answers include numbered citations.
7. Google Search Central, AI features and your websiteIndexed and snippet-eligible; no additional requirements.
8. Google Search Central, optimizing for generative AI searchQuery fan-out; ignore llms.txt and inauthentic mentions; Generative AI performance report.
9. Google Search Central blog, Top ways to ensure your content performs well in Google's AI experiences on SearchPreview controls, structured data matching visible content (May 2025).
10. Apple, App Store searchTitle, subtitle, keywords and primary category feed text relevance.
11. Aggarwal et al., GEO: Generative Engine Optimization (arXiv 2311.09735)Citing sources, quotations and statistics helped most; keyword stuffing did not.

Related AI search guides