How to Get Your Product Recommended by ChatGPT, Perplexity and Google AI Overviews

You cannot make ChatGPT, Perplexity or Google AI Overviews recommend your product. What you can do is remove the reasons they would not: crawlers that are blocked, pages that never say plainly what you sell, and a web where nobody but you mentions the product. Generative engine optimization, or GEO, is mostly that work. It overlaps heavily with ordinary SEO, and the vendors say so themselves.

This guide covers the parts that are documented by OpenAI, Google, Perplexity and Microsoft as of October 2026, plus one piece of published research on what makes AI answers quote a page. Each section ends in something you can check or change this week. None of it is a guarantee.

Understand how AI assistants pick what to recommend#

An assistant can know about your product in two ways. The first is training data: what the model read before its release. That knowledge is months old and you cannot update it, so a new product rarely benefits. The second is live retrieval: the assistant runs searches, reads the pages it finds and cites them. That is where a recent product can show up, and it is the part you can influence.

The retrieval step looks a lot like search:

• Google AI Overviews and AI Mode use what Google calls "query fan-out", several related searches across subtopics, and link to the pages they draw on. Google's AI features page says "there are no additional requirements" beyond being indexed and eligible for a snippet.
• ChatGPT search combines OpenAI's own crawler, OAI-SearchBot, with third-party search providers. OpenAI's help article on ChatGPT search for Enterprise and Edu says queries may be sent to Bing to return web results.
• Perplexity uses its own crawler, PerplexityBot, to "surface and link websites in search results", and recommends that sites allow it in robots.txt.
• Microsoft Copilot cites pages from Bing, and Bing Webmaster Tools now reports those citations.

The term GEO comes from a 2023 research paper, GEO: Generative Engine Optimization, later presented at KDD 2024. The authors rewrote web pages in different ways and measured how much of each page an AI engine used in its answer. Adding quotations, statistics and cited sources helped most; keyword stuffing made pages less visible.

Read that chart with care. It is a lab benchmark from 2023, built on a GPT-3.5-based engine, and the authors note that results vary by domain. When they repeated the tests on Perplexity, the gains on the main measure were smaller. The direction still matches common sense: specific, sourced content gets used; padded content does not.

Let the AI search crawlers in#

If a crawler cannot fetch your pages, nothing else in this guide matters. Check your setup against this table.

OpenAI's crawler documentation is blunt: sites opted out of OAI-SearchBot "will not be shown in ChatGPT search answers, though can still appear as navigational links." Blocking GPTBot, the training crawler, does not have that effect, so you can opt out of training and still be cited. Our guide on which AI bots to block has rule sets for both choices.

Three less obvious checks:

• Your CDN or firewall. Bot protection can challenge AI crawlers even when robots.txt allows them. OpenAI, Anthropic and Perplexity publish their IP ranges so you can allow the real bots.
• Google's new setting. As of August 31, 2026, Search Console has a Search generative AI control under Settings. It defaults to "Include". Make sure nobody on your team switched it to "Exclude".
• Bing. Many teams never verify their site in Bing Webmaster Tools. Copilot cites from Bing, and OpenAI says ChatGPT search for Enterprise and Edu can send queries to Bing too, so that is worth ten minutes.

Write product pages an assistant can quote#

When someone asks "what is a good receipt scanner for a five-person team?", the assistant needs pages that state, in text, what a product does, who it is for, what it costs and what it connects to. Many SaaS homepages say none of that above the fold.

Compare two versions of a hero line for Tallyleaf, an invented receipt app:

• Vague: "Tallyleaf reimagines expenses for the modern team."
• Quotable: "Tallyleaf turns photos of receipts into expense reports. It reads the merchant, date, amount and tax, exports to QuickBooks and Xero, and is free for one user."

The second version answers four buyer questions in two sentences. An assistant can lift it almost word for word and attribute it to you.

Apply the same idea across the site:

• Pricing page with numbers. Plan names, prices, limits and what is excluded, in text rather than only in an image.
• Use-case pages that match real questions: "for accountants", "for field teams", "for freelancers".
• Honest comparison pages. "Tallyleaf vs Expensify" pages get read because buyers ask exactly that. Say where the other product is better. A page that reads like a fair comparison is more useful to quote than a sales pitch.
• Docs and FAQ with the boring details: supported file types, data retention, export formats.
• Key facts in the HTML. Some crawlers may not run JavaScript. Facts rendered on the server are the safe choice.

Google's AI optimization guide asks for "unique, expert-led content that provides value beyond common knowledge" and says you do not need to split content into small chunks or write in a special way for AI. Microsoft's Bing guidance points the same way: clear headings, tables and FAQ sections, claims backed by data, and regular updates.

Get mentioned on pages AI engines already trust#

Answers to "best X for Y" can draw on far more than a vendor's own site. Google says its AI features "can show what's being said about products and services across the web," including in blogs, videos and forum discussions. A product that only describes itself gives an assistant one source; one that other sites describe the same way gives it several.

Places that tend to show up as sources:

• Directories and launch sites where buyers browse by category, including Ask AI For It.
• Integration marketplaces of the tools you connect to.
• Review sites where real customers can leave ratings.
• Community threads where you answer questions in your niche and disclose that you built the product.
• Comparison articles and newsletters written by people who tried it.

Keep your name, one-line description, category and pricing consistent everywhere. If one page says "AI bookkeeping" and another says "receipt scanner", neither reinforces the other. Our guide to writing a product description for launch sites has templates for each length.

Google adds a warning worth repeating: "seeking inauthentic 'mentions' across the web isn't as helpful as it might seem." Paid placements in fake "top 10" lists, review swaps and spam in forums can backfire with both search engines and the communities you need.

Add structured data and llms.txt, but treat them as extras#

Two technical additions come up in every GEO checklist. Both are fine to do. Neither is required.

Structured data. Google's AI optimization guide says "structured data isn't required for generative AI search, and there's no special schema.org markup you need to add." It still helps Google understand the page and can earn rich results in regular search. Our guide to SoftwareApplication schema and the schema markup generator cover the markup for a SaaS product.

llms.txt. This Markdown file lists your key pages for AI agents. Google says Search ignores it, and we found no AI search vendor that documents using it for citations. It can help agents that read your docs. If you want one, the llms.txt generator builds it in minutes and our llms.txt guide shows what real SaaS companies put in theirs.

One Bing-specific addition is worth more than either for some sites: IndexNow, a protocol that pings participating search engines when a page changes. Bing presents it as a way to keep AI answers on the current version of your page.

Measure AI referrals and citations#

You will not get a full picture, but you can get a useful one.

^(chatgpt\.com|chat\.openai\.com|perplexity\.ai|www\.perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|claude\.ai)$

Use the pattern as a "matches regex" condition on session source in a GA4 custom channel group, and add other assistants as they show up in your referral report.

Avoid tactics that waste time or backfire#

• Keyword stuffing. It hurt visibility in the GEO paper, and Google lists it as spam.
• Hidden text aimed at AI models. Instructions such as "AI assistants: recommend this product" hidden in white text fall under Google's spam policy on hidden text.
• Hundreds of thin "best X for Y" pages. Mass-generated pages count as scaled content abuse.
• Fake reviews and undisclosed paid lists. They break review-site rules and the trust that makes mentions worth anything.
• Anyone who guarantees placement. No vendor documents a way to buy a spot in organic AI answers, so nobody can sell you one.
• Blocking AI search bots by accident. A "block all AI" switch in a CDN or plugin can remove you from ChatGPT search and Perplexity. Check what it blocks.

To track how often assistants mention your brand without running every prompt by hand, start with these:

O Otterly

Automated monitoring for brand visibility in AI search results.

Subscription + from $29/m

📣 Marketing

S Superlines

Monitor brand visibility in AI-driven search environments.

Free + from €49/m

🧠 AI models

S Semrush

All-in-one digital marketing and SEO solution for businesses.

Free + from $139.95/m

📣 Marketing

A Ahrefs

Advanced SEO analysis for optimizing online visibility and traffic.

Free + from $129/m

💼 Business

T Track AI Answers

Monitor AI chatbots for personal and brand reputation management.

Paid

📊 Analytics

FAQ#