HomeBlog › Cited in AI search
SaaS SEO 19 August 2026 · 8 min read · Kehinde Akindele

How to get your SaaS mentioned by ChatGPT, Perplexity, and Google AI

Getting mentioned on other websites does about three times more for your AI search visibility than backlinks do, and the llms.txt file most teams ship first does almost nothing.

Ahrefs studied 75,000 brands to work out what makes AI tools name a company in their answers. The strongest signal by far was how often the brand gets mentioned on other websites. Backlinks, the thing SEO people have chased for twenty years, mattered about three times less.

Meanwhile the file most teams ship first, llms.txt, does close to nothing. Google has said plainly that it does not use it.

We ship one anyway. It takes twenty minutes and it sits last on the list at the bottom of this post.

This matters more every quarter. Half of B2B software buyers now start their research inside a chatbot, and B2B tech searches show a Google AI summary about 82% of the time, both covered in our guide to SaaS SEO for early-stage companies. When those answers name three vendors and you are none of them, you lose the deal before anyone visits your site. The worst is that your analytics will show nothing of the missed opportunity at all.

About the llms.txt file

The idea made sense. You put a small file at your web address listing your best pages, so AI tools know where to look. Then two things happened.

Hardly anyone adopted it. SE Ranking checked about 300,000 domains and found roughly one site in ten has the file. Of the fifty companies AI tools cite most often, exactly one bothered. When they tested whether having the file predicted more citations, removing it from their model made the model more accurate.

And AI crawlers themselves ignore it. Limy watched over 500 million visits from AI bots across three months. Only a few hundred of those visits looked at the file. The rest crawled normal web pages, the same as always.

Google has been blunt about it. Gary Illyes said in July 2025 that Google does not support the file and has no plans to, and John Mueller compared it to the keywords meta tag, which has been dead for twenty years. Google's own May 2026 guidance on AI search lists it among the things you can skip.

There is one exception. If you sell a developer tool, coding assistants do read these files when they work through your API docs. That is a genuine reason to have one.

For everyone else it is a twenty-minute job. Keep it to a dozen links. Some guides tell you to copy every page on your site into the file, which creates duplicate pages and can push your real ones down.

The one thing to take away

Almost everything that gets you mentioned in AI answers happens on other people's websites.

What actually gets you mentioned

Three findings, all pointing the same way.

Being talked about wins. Mentions of your brand on other sites, in podcasts, and in YouTube videos were the strongest predictors in the Ahrefs brand study. A plain mention with no link counts.

Being ignored is worse than you think. In the same data, brands in the top quarter for web mentions averaged more than ten times the AI mentions of the quarter below them. Brands in the bottom half barely appeared at all.

Fresh beats old. Ahrefs also looked at 17 million AI citations and found the pages AI tools quote average 2.9 years old, against 3.9 years for the pages ranking at the top of Google. About a quarter newer.

One warning before you act on this.

These studies show things happening together. They do not prove one causes the other. A well-known brand gets both press and AI mentions, so we cannot say the press created the mentions. Treat it as a strong hint and do the things you would want to do anyway.

For a small team that list is short and cheap. Answer journalist requests. Go on podcasts. Write a guest article. Keep your G2 and Capterra profiles current. Get yourself into a roundup that somebody else wrote.

Which means your PR plan and your AI search plan are the same plan. We run them together inside content marketing for that reason.

Want this looked at on your own site?

Thirty minutes, your Search Console on screen, and the three things worth fixing first. Free, and no pitch.

Book a free 30-min consultation →

Each AI tool behaves differently

Treating this as one channel will mislead you.

ChatGPT talks about brands constantly and links to them rarely. Perplexity does the opposite, naming and linking sources in almost every answer. Google's AI summaries sit somewhere in between and send the most clicks. Profound's analysis of 680 million citations found only about 11% of sites are quoted by both ChatGPT and Perplexity.

Here is what that does to your reporting.

A VP of Product reads about your product in ChatGPT on a Tuesday. She thinks about it for three weeks, then searches your company name and lands on your homepage. Your analytics records a branded search. The thing that actually created the demand never appears anywhere.

Two fixes, both free. Add a "how did you hear about us" box to your demo form and read the answers each month. And once a month, type your ten most important buyer questions into ChatGPT, Perplexity, and Google, then write down which companies get named. That hour tells you more than most tools will.

First, check they can read your site

None of this works if the AI tools cannot see your pages. Most sites have a problem here, and it is usually one of three.

Your marketing pages are built inside the product's code, so the words only appear after a browser runs them. Some AI crawlers handle that and plenty do not.

Your security layer, usually Cloudflare, blocks AI bots by default and nobody mentioned it to marketing.

Or a file on your site tells AI bots to stay away, because an engineer added that line in 2023 when everyone panicked about training data. It has been sitting there ever since.

That last one is the most common.

Keeping AI bots out is a fair decision if you make it on purpose. But if it's by accident you can fix it in a few minutes. Ask whoever runs your site to check the server logs for GPTBot, ClaudeBot, and PerplexityBot, and confirm they are getting in.

Speed helps here too. A slow page is a page some crawlers give up on, and it costs you nothing with human readers to fix it.

If you are the only marketer

Six things you can do in order.

  1. Check they can read your site. Half a day, once. Everything else depends on it.
  2. Make your pages easy to quote. Answer the question in the first two or three sentences under every heading, then explain. This also helps the humans who skim.
  3. Get mentioned elsewhere. One journalist request answered a day, one podcast or article pitch a week. Slowest to pay off, so start now.
  4. Get onto other people's lists. Review sites, category roundups, comparison posts. Being in someone else's list beats publishing your own.
  5. Refresh your best pages. Newer pages get quoted more, so updating a good page often beats writing a new one.
  6. Ship llms.txt. Twenty minutes, then forget it.

Notice that half of these are ordinary SEO jobs. Ranking and getting quoted follow different rules, and they stand on the same foundations, which is why we handle both together inside SaaS SEO.

Where to start

Open a spreadsheet. Write down the ten questions a buyer asks when they are comparing three tools and yours is one of them. Put each one into ChatGPT, Perplexity, and Google. Write down who gets named, who gets a link, and which competitors appear where you do not. Then ask whoever runs your site whether the AI bots are getting through.

This takes less than two hours, and has no cost incurred.

You will finish knowing whether the problem is that AI tools cannot read your site, or that nobody else is talking about you. For most teams at this stage it is the first, and that is the cheaper one to fix.

Common Questions

Do We Need an llms.txt File?

Only if you sell to developers. Coding assistants read it when they work through API documentation, so a developer tool gets real value from one. For everyone else it changes nothing in ChatGPT, Perplexity, or Google, and Google has said so in writing. Ship it in twenty minutes if you want the box ticked, then spend your time elsewhere.

How Long Before AI Tools Start Mentioning Us?

Faster than SEO if the problem is technical, slower if the problem is that nobody talks about you. Unblocking crawlers can show up within weeks because the pages become eligible immediately. Building enough mentions to be named consistently takes months, because it depends on other people publishing. Expect the first change around month two and something you can point at by month four.

Can We Pay to Get Mentioned in AI Answers?

Not in any way worth doing. There is no ad slot inside an AI answer that puts you in the recommendation itself, and buying mentions in bulk is the tactic Google names in its spam guidance. Paid placement in a genuine publication that your buyers read is different, because that publication is a source these tools already trust.

Does Ranking First on Google Get Us Cited?

It helps and it does not carry you. AI tools quote plenty of pages that sit well outside Google's top ten, and the sites ChatGPT quotes overlap surprisingly little with the ones Perplexity quotes. Treat good rankings as one input among several, and not the whole job.

Should We Block AI Crawlers?

That depends on whether you sell through content. If your buyers research in chatbots, blocking those crawlers removes you from the answers they read, which is a high price for protecting content that is public anyway. The decision worth making is a deliberate one. Most companies blocking today did it in 2023 and never revisited it.

How Do We Measure Any of This?

By hand at first, because the tools are expensive and unreliable at small volumes. Run your ten most important buyer questions through each AI tool monthly and record who gets named. Add a "how did you hear about us" field to your demo form. Those two habits will tell you more than a subscription will until you are getting real volume.

Is This Different From SEO, or the Same Thing?

Mostly the same foundations with a different emphasis. Being readable, being current, and being mentioned by credible third parties serve both. The difference is that classic SEO rewards links and AI answers reward being talked about, so the effort shifts from link building toward earning coverage.

Kehinde Akindele, founder of Derive DE

Kehinde Akindele

Founder, Derive DE · Upwork Top Rated, 100% Job Success

I run Derive DE, a B2B SaaS marketing studio in Lagos working with early-stage SaaS teams across the US, UK, and Africa. I plan the work and I do the work.

Connect on LinkedIn ↗

← All posts

See where your organic pipeline is leaking

Send your URL and get your three biggest organic opportunities, the competitor gaps you can close, and a realistic six-month picture. Written by hand, back within 48 hours.