GEO for Indie Founders: Getting Cited by ChatGPT, Claude, and Perplexity

A growing share of the people looking for a tool like yours now ask an assistant instead of a search engine, and the assistant names three or four products. Generative engine optimization, or GEO, is the work of being one of them. Most of what is written about it is speculation sold as method, so this post separates the parts that are cheap and certain from the parts nobody outside OpenAI and Anthropic actually knows.

Table of contents

• What GEO is, and how it differs from SEO
• Step one: let the crawlers in
• Write sentences that can be quoted
• llms.txt, honestly
• The part that actually moves the needle
• Why Bing matters more than you think
• How to tell whether it is working
• Frequently asked questions

What GEO is, and how it differs from SEO

SEO competes for a position in a list of links. GEO competes to be named inside an answer. The difference matters because of what the two formats do to your traffic. A search result gives you a click. An AI answer often gives you a mention with no click at all, and the reader arrives later by typing your name directly, if they arrive at all.

That has two consequences worth internalising before you spend time on any of this. Your analytics will understate it, because a recommendation that turns into a direct visit three days later looks like direct traffic. And being named once in the right answer is worth more than ranking tenth for a term nobody clicks, because the answer is the whole page.

The useful mental model: the assistant is summarising a rough consensus it formed from many pages about your category. You are not optimising one page. You are trying to be part of what the internet appears to agree on.

Step one: let the crawlers in

This is the only part of GEO that is genuinely mechanical, and it takes ten minutes. AI companies run named crawlers, and a surprising number of sites block them by accident, usually because a security plugin or a hosting default did it. Check that these are allowed in your robots.txt:

• GPTBot and OAI-SearchBot, OpenAI's training and search crawlers
• ClaudeBot, Anthropic's
• PerplexityBot
• Google-Extended, which governs Gemini's use of your content separately from normal Google indexing

Lighthouse takes the simplest possible position here: one wildcard rule allowing everything, plus a sitemap. There is no per-crawler allowlist to maintain and no chance of forgetting a new bot when one appears.

User-Agent: *
Allow: /

Sitemap: https://lighthouse.build/sitemap.xml

Blocking these crawlers is a legitimate choice if you sell your writing. If you sell software, blocking them means opting out of being recommended, which is a strange trade to make on purpose.

Write sentences that can be quoted

This is the highest-leverage on-page change, and it is a writing problem rather than a technical one. Assistants lift short, factual, self-contained statements. They struggle with marketing copy, because marketing copy asserts feelings rather than facts.

Three practical rules. Put the plain description in the first two sentences of your homepage, your about page, and any file you write for machines, because that is what gets extracted. Include the actual price as a number, since pricing is the single most common thing people ask an assistant about a tool. And keep those sentences current: stale claims are worse than none, because the assistant will repeat a price or a plan that no longer exists long after you changed it.

llms.txt, honestly

llms.txt is a proposed file, in the spirit of robots.txt, that gives language models a clean summary of your site and links to the pages that matter. Lighthouse publishes one, so this is not a criticism from the sidelines.

It is also worth being straight about its status, because most posts on this topic are not. It is a community proposal rather than an adopted standard, and no major AI company has publicly committed to reading it. Anyone telling you it is how you get into ChatGPT is guessing.

So why publish one? Because it takes twenty minutes, it costs nothing to serve, and the exercise of writing it is valuable on its own. Compressing your entire product into a few factual sentences and a list of your best pages is exactly the raw material described in the previous section. If engines do start reading it, you are ready. If they never do, you still improved the copy you use everywhere else.

Keep it short and factual: what the product is, who it is for, what it costs, then links to your core pages, docs, and best articles. Set a reminder to refresh it when pricing or plans change.

The part that actually moves the needle

Everything above is table stakes and none of it is a differentiator, because your competitors can do all of it in an afternoon. The thing that decides whether an assistant names you is what other sites say about you. The model is summarising a consensus, and one site describing itself is not a consensus.

In rough order of weight:

• Review platforms. A handful of honest reviews on the sites people use to compare software changes the answer to "best X tool" more than anything you write yourself. Never buy or fake these. Ask real users, accept the ones you get.
• Alternative and comparison listings. Being listed as an alternative to the incumbents feeds every "alternatives to X" question, which is one of the most common shapes of AI query in software.
• Directories with real pages. A listing page that describes your product in someone else's words is a small, independent statement about what you are. Their value is being third-party, not being a link.
• Mentions in other people's writing. A blogger including you in a roundup is worth more than three pages you wrote about yourself.
• Your own comparison pages. Genuinely useful, and they work best when honest about what the competitor does better. An assistant reading a one-sided comparison has no reason to trust it.

One caution learned the hard way: some of these listings are one person's side project on a free hosting plan, and they disappear. When that happens the page, the description, and the link all vanish together. Spread the effort across many rather than counting on any one of them. I went through which of these are worth the time in which launch directories are actually worth it.

Why Bing matters more than you think

ChatGPT's search leans on Bing's index rather than Google's, which means Bing Webmaster Tools is a more direct lever on AI visibility than it has ever been on ordinary traffic. It is free, it takes about fifteen minutes, and almost nobody building a small product bothers.

Verify your domain, submit your sitemap, and then leave it alone. It is not a channel you work, it is a box you tick once so that the index feeding a major assistant knows your pages exist.

How to tell whether it is working

There is no rank tracker for this, and the honest answer is that measurement is crude. Three things are worth watching.

1. Crawler activity

Your server logs, or a crawl report from an SEO tool, will show the named bots arriving. When I checked Lighthouse's, ChatGPT's crawler had visited ten times in a single day. That is a real signal that the door is open and the site is considered worth re-reading, and it is the earliest feedback you will get. It is not evidence of being cited. Crawling and citing are different things.

2. Ask the assistants

Keep a short list of the questions a potential customer would actually ask, and run them monthly across two or three assistants. For a waitlist tool that list looks like "best waitlist tool for an indie SaaS launch", "alternatives to the market leader", and "what is Lighthouse". Record whether you were named and whether the description was accurate. Wrong descriptions are the most actionable result you can get, because they point at a specific stale sentence somewhere.

3. Direct traffic and how people arrive

Because a recommendation frequently converts into a direct visit later, watch for direct traffic growing without a matching campaign. It is a weak signal on its own. Paired with the other two it is usually the closest thing to proof you will get.

Frequently asked questions

Is GEO replacing SEO?

No, and the framing is a trap. Nearly everything that makes you citable also makes you rank: clear writing, factual pages, other sites referencing you. The difference is emphasis rather than method. Treat GEO as a lens on the same work, not a second budget.

Should I pay for a GEO service or tool?

At the stage where you are still validating an idea, no. The mechanical parts take an afternoon and the rest is earning genuine third-party mentions, which no tool does for you. Be especially wary of anyone guaranteeing citations or selling packages of links, which is the same scheme that has been sold to founders for twenty years under a new name.

How long does it take?

Longer than SEO in one sense and shorter in another. The models update their picture of a category slowly, so nothing you do today shows up tomorrow. But a new category with few competitors can be entered quickly, because there is not much consensus to displace.

Does structured data help?

It is cheap and it certainly does not hurt. Describing your product, price, and organisation in schema gives a machine an unambiguous version of the facts you are also stating in prose. Do it once, then spend your remaining effort on the consensus problem.

My product is brand new. Is any of this worth doing yet?

Do the ten-minute parts now, because they compound and cost nothing: open robots.txt, plain factual sentences, Bing, a couple of honest listings. Skip everything else until you have users, since reviews and mentions are the real lever and you cannot manufacture either.

The uncomfortable summary is that GEO rewards the same things good products have always been rewarded for, with less room to game it. Say plainly what you do and what it costs, let the crawlers read it, and get enough real people talking about you that a summary of the internet includes your name. The technical checklist is an afternoon. The rest is the same work as building a business.

Lighthouse is a waitlist, survey, newsletter, and feedback toolkit for indie founders, at $19 or $29 per month. From an indie dev, for indie devs and makers.