How to get your SaaS recommended by ChatGPT, Claude, Perplexity and Gemini

The short answer: AI assistants recommend the products that the pages they read already recommend. To get named, make sure assistants can reach your site, publish pages that directly answer the questions your buyers ask, and get mentioned on the third-party pages assistants quote: review sites, comparison articles, forums and videos. Then measure, every week, how often you're named across the same set of questions.

The rest of this guide explains how that works in practice, step by step.

How AI assistants decide which software to recommend

When a buyer asks "What's the best expense software for a 50-person startup?", an assistant answers in one of two ways.

From memory. The model answers from what it learned in training. You can't influence this quickly: it reflects what the web said about your category months or years ago, and a new product may simply be absent.

From a live search. ChatGPT search, Perplexity, Gemini, Claude with web search and Google's AI Overviews look things up before answering. OpenAI says ChatGPT search "leverages third-party search providers, as well as content provided directly by our partners", and Google says its AI features may use "query fan-out", running several related searches across subtopics before writing one answer.

The live-search mode is where you can win, and win fast. The assistant reads a handful of pages, decides which products those pages agree on, and names two or three of them. So the job is to be on the pages it reads, and to be the clearest answer on your own site.

Which pages get read varies a lot by assistant and by month, so treat any single number with caution. Profound's study of 680 million citations found Wikipedia made up 47.9% of ChatGPT's top-10 cited sources and Reddit 46.7% of Perplexity's between August 2024 and June 2025. Ahrefs found YouTube (22.9%) and Reddit (18.5%) led the domains cited in Google AI Overviews in 2026. Semrush saw Reddit's presence in ChatGPT answers fall from close to 60% to around 10% after September 2025. The lesson isn't "post on Reddit"; it's that assistants lean on independent sources, and which ones changes over time.

Step 1: Write down the questions your buyers actually ask

You can't optimise for "AI" in general. You optimise for specific questions. Pick 10 that a real buyer in your category would type, across the stages of a purchase:

• Discovery: "What's the best [category] for [type of company]?"
• Use case: "[Category] that works with [tool you integrate with]"
• Comparison: "[Your competitor] vs [another competitor]" and "alternatives to [market leader]"
• Evaluation: "Is [your product] good for [use case]?" and "How much does [category] cost?"

Pull wording from sales calls, support tickets and the categories buyers browse on review sites. Questions about specific situations ("for a 20-person agency", "for clinics in the UK") matter most: that's where challengers can win against bigger brands.

Step 2: Measure your baseline before you change anything

Ask each of your 10 questions in ChatGPT, Claude, Perplexity, Gemini and Google (for AI Overviews). That's 50 answers. For each one, note:

1. Whether you're named, and in what position.
2. Which competitors are named instead.
3. Which sources the answer cites or links to.
4. Anything the assistant gets wrong about you: old pricing, missing features, the wrong category.

Use web search mode where an assistant offers a choice, start a fresh chat for every question, and run your most important questions twice. Answers vary from one run to the next, so a single check can mislead you. The count you end with ("named in 4 of 50 answers") is your scoreboard.

The sources list is the most useful part. It tells you exactly which pages you need to appear on.

Step 3: Make sure assistants can reach your site

It sounds obvious, but blocked crawlers are a common reason a brand never appears. Each company uses separate crawlers for training, for search and for fetching a page when a user asks, and blocking the wrong one can quietly remove you from AI answers.

Three things to check:

• Your robots.txt. If you block OAI-SearchBot, OpenAI says your site won't be shown in ChatGPT search answers. Blocking the training crawlers (GPTBot, ClaudeBot) is a separate choice and doesn't remove you from search answers.
• Your CDN or firewall. Some bot-protection settings block AI crawlers by default. Check your Cloudflare or hosting settings, not only robots.txt.
• Normal search indexing. For AI Overviews, Google says a page only needs to be indexed and eligible to show with a snippet. Submit your sitemap in Google Search Console and Bing Webmaster Tools; OpenAI uses third-party search providers, and Microsoft announced Bing as one in 2023.

You don't need special files or markup for Google. Its own guidance says optimising for generative AI search "is still SEO", and that there's no special schema to add.

Step 4: Publish pages that answer buyer questions directly

Most SaaS homepages say something like "The all-in-one platform for modern teams." An assistant can't match that to "expense software for a 50-person startup." It needs a page that says, plainly, who you're for and what you're best at.

For each of your 10 questions, make sure one page on your site gives a direct, quotable answer:

• Use-case pages: "Expense management for startups of 20–200 people", with the specifics a buyer cares about.
• Comparison pages: "[You] vs [competitor]" that are fair, specific, and include where the competitor is the better choice. One-sided pages get ignored by readers and by assistants.
• "Alternatives to [leader]" pages: buyers ask this constantly, and assistants look for pages that answer it.
• A clear pricing page, in text, not only in images or behind "contact us". Assistants repeat whatever price they find, including old ones.

How you write matters too. The research paper that coined the term GEO tested nine ways of rewriting pages and found that adding statistics and adding quotations from credible sources improved visibility in AI answers the most, by up to 40% overall. Citing sources and improving readability also helped. Keyword stuffing did not.

In practice:

1. Put the answer first. Open each page or section with a two- or three-sentence answer, then the detail.
2. Use specific numbers. "Approvals take under 2 minutes on average" beats "fast approvals". Only use numbers you can stand behind.
3. Quote and cite. A line from a customer, an analyst or your own data, with the source.
4. Use headings phrased like questions, so an assistant can find the right part of the page.

Step 5: Get mentioned on the pages assistants already quote

This is where most of the movement comes from. Go back to the sources list from step 2. For each page an assistant cited:

• "Best [category]" articles: check whether you're listed. If not, contact the author with a short, useful note: what you do differently, a customer example, a screenshot. Prioritise independent publishers over competitor-run lists.
• Review sites: a complete, current profile with recent reviews from real customers. Ask happy customers to review you; never buy or fake reviews.
• Forums and communities: take part in Reddit and industry communities where your buyers ask questions. Answer genuinely, disclose that you work for the company, and don't drop links where they aren't wanted.
• YouTube: short, specific videos ("How to set up approval rules in [product]") get cited by Google's AI features.
• Partners and integrations: your integration partners' marketplace pages and blogs are trusted sources that already rank.

Step 6: Fix what AI gets wrong about you

Assistants often repeat outdated facts: an old price, a feature you've since shipped, an integration you no longer offer. They get these from somewhere, usually an old article, a review-site profile or a page on your own site that was never updated.

Trace each wrong fact back to its likely source, update or correct it, and make sure your own site states the current fact clearly in text. Then use the same one-sentence description of your company everywhere: your site, review profiles, LinkedIn, directories. Consistency helps assistants recognise you as one entity.

Step 7: Track it every week

Re-run the same 10 questions across the same five assistants every week, the same way. Keep a simple sheet: date, question, assistant, named or not, position, sources. Over a quarter the trend will be clear even though individual answers wobble.

Two extra signals:

• Referral traffic. ChatGPT adds utm_source=chatgpt.com to the links it sends, so you can see that traffic in your analytics. Perplexity and others show up as referrers.
• Sales conversations. Add "How did you hear about us?" with an "AI assistant" option. Many buyers who found you through AI never click a link.

What doesn't work

• Keyword stuffing. The GEO research found it offered little to no improvement.
• Relying on llms.txt. It's a 2024 proposal for a file that summarises your site for AI tools. It's harmless to add, but Google says its search doesn't use it, and no major AI company has confirmed using it for answers.
• Publishing lots of thin, AI-written articles. Assistants quote sources that add something new. Ten useful pages beat a hundred generic ones.
• Fake reviews or paid placements disguised as editorial. They're against review-site rules and can backfire badly.

How long it takes

With live-search assistants, changes can show up within weeks of the pages they read changing: once your new pages are indexed and a few third-party pages mention you. Google AI Overviews tends to follow your normal Google rankings, which take longer for a new site. The model's built-in memory only updates with new model versions, so expect that part to lag by months. These are typical patterns, not guarantees; your category and starting point matter a lot.

Frequently asked questions

Can I pay to be recommended by ChatGPT?

You can't pay for a place in ChatGPT's organic answers. OpenAI has started showing clearly labelled ads in ChatGPT in some markets, but those are separate from the recommendations inside answers. Organic recommendations depend on what the sources the assistant reads say about you.

Does schema markup help me appear in AI answers?

Google says there's no special schema.org markup needed for its AI features. Structured data can still help search engines understand your pages, but it won't get you recommended on its own. Clear content and third-party mentions matter far more.

Why does ChatGPT give different answers each time?

Answers are generated fresh each time and can draw on different search results, so the products named can change between runs. That's why you should measure across many questions and runs, and watch the trend rather than any single answer.

Is this different from SEO?

It overlaps heavily. Being indexed, ranking well and having clear pages all help. The differences are what you measure (being named in answers, not just clicks) and where you focus: third-party mentions and direct, quotable answers carry more weight.

How do I know which questions to track?

Start with the questions your buyers ask in sales calls, then add "best [category] for [segment]", "[competitor] alternatives" and "[you] vs [competitor]". Ten well-chosen questions are enough to start.