Science of Apps - Launchpad · Explainer

How AI engines
recommend products

AI engines recommend products by searching the web when they answer, reading the pages their search bots can reach, and summarising what those sources agree on, with links. Google adds query fan-out across subtopics. So a product gets recommended when crawlable pages answer the buyer's specific question and independent sources describe it consistently. None publishes a ranking formula.

By , FounderUpdated 5 min read

Share

Quick answer: How do AI engines decide which products to recommend?

  1. Search at answer time: ChatGPT search, Claude, Perplexity and Google's AI features fetch web pages.
  2. Bots decide reach: OAI-SearchBot, Claude-SearchBot, PerplexityBot, Googlebot.
  3. Fan-out: Google runs related searches across subtopics for one answer.
  4. Sources: answers link the pages they used.
  5. Agreement: consistent facts across independent sites are easier to state confidently.
  6. Formula: none of them publishes one; placement isn't guaranteed.

Key takeaways

  • An engine can't recommend what its search bot can't reach.
  • Specific buyer questions get specific answers; generic pages rarely make the list.
  • Independent sources matter: your own site is one voice.
  • OpenAI says placement in ChatGPT search isn't guaranteed even when you allow its bot.
  • Inauthentic mentions don't help much, says Google; real coverage does.

"What's the best tool for…?" is now a question people ask an AI assistant before they ask Google. The answer is a short list with links, and founders want to know how products get onto it. None of the engines publishes a formula, but each documents enough about how it searches and reads the web to explain the mechanics. This explainer puts those pieces together from OpenAI's, Google's, Anthropic's and Perplexity's own documentation, read on 17 November 2026. Jump to the pipeline, what makes a product recommendable or how to check.

The pipeline: from question to recommendation

Diagram of 5 steps for the pipeline: from question to recommendation: 1. The question; 2. Search; 3. Fetch; 4. Summarise; 5. Link.
The pipeline: from question to recommendation
  1. The question — "best invoicing app for freelancers in India".
  2. Search — the engine searches the web; Google may fan out into related sub-searches.
  3. Fetch — its search bot reads the pages it can reach.
  4. Summarise — it states what the sources say, often a short list.
  5. Link — it cites the pages it used.
EngineBot that reads your pagesWhat the company says
ChatGPT searchOAI-SearchBotAllow it to appear; "placement is not guaranteed"
Google AI Overviews and AI ModeGooglebotIndexed and snippet-eligible; may use query fan-out
ClaudeClaude-SearchBot, Claude-UserBlocking them may reduce visibility
PerplexityPerplexityBot, Perplexity-UserAllow it to appear in results; not used for training

Table: How AI engines reach your pages (each company's documentation, checked 17 Nov 2026)

What makes a product recommendable

Checklist: What makes a product recommendable — Reachable; Specific; Factual; Corroborated; Genuinely mentioned; Current.
What makes a product recommendable
  • Reachable — the engine's search bot can fetch your pages
  • Specific — a page answering the exact buyer question, not a generic homepage
  • Factual — prices, features, who it's for, dated
  • Corroborated — independent sites describe you the same way
  • Genuinely mentioned — Google says inauthentic mentions aren't as helpful as they seem
  • Current — stale facts get dropped or repeated wrongly

The GEO study (KDD 2024) found that adding statistics, quotations and citations raised a source's visibility in generative engines, while keyword stuffing did poorly.

Why your competitor gets recommended

Likely reasonFix
Your pages block or challenge the botAllow the search bots; check the firewall
No page answers that exact questionWrite one, answer first
Fewer independent sources describe youDirectories, profiles, genuine coverage
Inconsistent facts about youOne name, one-liner and price everywhere
Your facts are out of dateUpdate and date them

Table: Common reasons a competitor is recommended instead (editorial)

How to check

Checklist: How to check — Ten buyer questions; Mentioned or not; Which sources are cited; Referrals; Google.
How to check
  • Ten buyer questions — asked in each engine, monthly
  • Mentioned or not — and in what words
  • Which sources are cited — those are the pages to be on
  • Referrals — chatgpt.com, perplexity.ai, claude.ai, gemini.google.com in analytics
  • Google — Search Console's Generative AI performance report
Science of Apps Launchpad homepage with the headline Launch your product. Get it found.
The Science of Apps Launchpad: an independent profile engines can read.Image: Science of Apps · © Science of Apps · source

Ready to be recommendable?

How we know

Mechanics come from OpenAI's crawler and ChatGPT search pages, Google's AI features documentation and generative AI optimisation guide, Anthropic's crawler help article and Perplexity's crawler documentation, all read on 17 November 2026, plus the GEO paper (KDD 2024). The "why your competitor" table is our editorial reading; none of the engines publishes its selection formula.

Key terms, defined

  • Search bot: a crawler fetching pages for answers.
  • Query fan-out: running several related searches for one answer.
  • Citation: a link to a source used in an answer.
  • Corroboration: independent sources agreeing.
  • Generative engine: an AI system that writes answers from sources.

Frequently asked questions

How does ChatGPT recommend products?

When ChatGPT searches the web, it reads pages its search crawler (OAI-SearchBot) can reach and summarises them with links. OpenAI says allowing OAI-SearchBot lets your pages appear in search results, but placement is not guaranteed.

How do Google AI Overviews choose which products to mention?

Google says AI Overviews and AI Mode may use query fan-out, running related searches across subtopics and sources, and show supporting links from pages that are indexed and eligible for a snippet. It doesn't publish how links are chosen.

Can I pay to be recommended by AI engines?

Not in their organic answers; none of these companies describes a paid route into answers. Be wary of services promising placement.

Why does an AI engine recommend my competitor and not me?

Usually because more independent pages describe the competitor clearly for that question, or because the engine's bot can't reach your pages. Check access first, then publish specific answers and earn consistent coverage.

Do reviews and directories affect AI recommendations?

They're among the independent sources engines can read. Google's guidance notes its AI features can show what's being said about products across the web, while inauthentic mentions aren't as helpful as they seem.

How do I check if AI engines recommend my product?

Ask each engine the ten questions your buyers ask, monthly, and note whether you're mentioned and what's said. Track referrals from chatgpt.com, perplexity.ai, claude.ai and gemini.google.com in analytics.

Sources

  1. OpenAI — Overview of OpenAI crawlers (checked 17 Nov 2026)
  2. OpenAI Help Center — Searching the web with ChatGPT (checked 17 Nov 2026)
  3. Google Search Central — AI features and your website (checked 17 Nov 2026)
  4. Google Search Central — Optimizing your website for generative AI features on Google Search (updated 10 Jul 2026)
  5. Anthropic (Claude Help Center) — Does Anthropic crawl data from the web, and how can site owners block the crawler?
  6. Perplexity — Perplexity crawlers (checked 17 Nov 2026)
  7. Aggarwal et al. — GEO: Generative Engine Optimization (arXiv 2311.09735, KDD 2024)
  8. Checked November 17, 2026. Figures that change — fees, salaries, prices — are reviewed every six months; if something is out of date, tell us.
Pass it on

Know someone who'd enjoy this?

Send it to them — one tap, no sign-up. Good ideas travel best between friends.

Keep exploring

Where to go from here

Join free

Get the ideas, not the noise.

Sign up with Google and we’ll send you short UX psychology tips by email — why people tap, stall and leave, and what to do about it. Unsubscribe anytime. Prefer something to keep? Get the free Laws of UX guide as a PDF.

Get the free Laws of UX PDFFree · no spam · unsubscribe anytime
Free, forever

Launch day ends. Your listing shouldn't.

Be one of the sources engines read: a free, crawlable Launchpad profile with consistent facts, ranked by relevance and never by payment, plus a product test that checks your findability.

Free PDF · 74 pages · 2 editions

The Laws of UX, in one field guide

60 research-backed principles in 6 categories — what each one means, why it works on the human brain and how to use it on your own screens. Written by Sushant Singh, founder of Science of Apps, who has scaled consumer apps to lacs of users and spent 10+ years leading product across startups and digital businesses in the US, EU & India.

Pick your edition
Free PDF · 90 pages · 2 editions

100+ Subscription Paywalls, decoded

110 real paywalls from top apps, taken apart, rated out of 5 and decoded with the Laws of UX — plus the frameworks, copy and checklists to ship your own. By Sushant Singh, founder of Science of Apps.

  • 110 paywalls rated
  • 34 Laws of UX
  • 12 frameworks

Which paywall scored 4.8 out of 5, and which scored 2.6? It’s all inside.

Pick your edition
Free PDF · 188 pages · 2 editions

Unmistakable: the logo book

How the world’s greatest logos are made, chosen and kept — and how to make yours. 98 short lessons and 19 true stories for founders, designers and the curious. By Sushant Singh, founder of Science of Apps.

  • 98 lessons
  • 19 true stories
  • 60+ real logos

Why did a $35 sketch become the Swoosh, and why did Gap lose its nerve in six days? It’s all inside.

Pick your edition