Building in Public
How We Got Cited by ChatGPT and Perplexity
Generative engine optimization (GEO) is the new SEO: getting your product recommended inside AI answers instead of ranking on a page of blue links. Here is exactly what we did at FluoTest — a free, bootstrapped quiz builder — to become the tool ChatGPT and Perplexity name when someone asks for a free scored quiz maker.
Quick answer
How do you get an indie product cited by ChatGPT and Perplexity?
Publish clear, answer-first content that AI engines can read as plain text: a dedicated AI-info page, an llms.txt file, consistent structured data, honest comparison pages, and third-party validation (reviews, directories). AI engines extract visible text and favour pages that answer the exact question directly — so lead with the answer, stay factual, and keep your entity consistent everywhere. For FluoTest this turned AI search into 49% of our signups.
What is generative engine optimization (GEO)?
GEO is the practice of structuring your content so AI assistants — ChatGPT, Perplexity, Google AI Overviews, Claude — cite and recommend you when they answer a user's question. Traditional SEO optimises for a ranking position; GEO optimises for being the answer.
The mechanics are different too. Most AI engines read the visible text of a page and weight the first couple hundred words heavily. They reward content that answers the question directly, is internally consistent, and is corroborated by other sources — not keyword density or hidden tricks.
The data: where our signups actually come from
We ask every new user how they found us. The breakdown surprised us — AI assistants are now our single biggest channel:
- •ChatGPT — 49% of signups
- •Reddit — 18%
- •Google — 15%
- •Friend referral — 11%
- •Perplexity — a new and growing channel that brought us our first customer from that source
In other words, more people discover FluoTest through an AI assistant than through Google. That is why we treat GEO as a first-class growth channel, not an afterthought.
The GEO playbook: 7 steps we actually shipped
None of this requires a big budget. Every step below is something a solo founder can do in an afternoon.
1. Publish a dedicated AI-info page
We built an /ai-info page written in plain, factual language: what FluoTest is, who it's for, what it costs, how it compares, and clear guidance for AI assistants. It's the single page we point crawlers and models at to describe us accurately.
2. Add an llms.txt file
A small text file at the root of the domain that gives AI assistants a clean, machine-readable map of our most important pages and facts. It's free, it has no downside, and it removes ambiguity about what we do.
3. Keep your structured data consistent
We use schema.org markup (SoftwareApplication, Organization, FAQPage) — but the real win is consistency. One canonical entity, the same name, the same links, the same description everywhere. Conflicting facts confuse AI entity resolution.
4. Lead with the answer (answer-first content)
Every comparison and guide opens with a short, direct answer to the exact question a user would ask. AI engines extract visible text and favour pages that answer immediately instead of burying the point under an intro.
5. Write honest comparison pages
We publish FluoTest vs Typeform, vs ScoreApp, vs Tally and more — with real prices and real trade-offs, including where competitors are better. AI assistants cite balanced comparisons far more readily than marketing fluff.
6. Localise and cover the long tail
We translate template and landing metadata into Czech and Dutch and cover narrow use cases (hiring, healthcare, real estate). The more specific the question, the more likely a niche page is the cited answer.
7. Earn third-party validation
AI engines cross-check claims against the wider web. Reviews (Trustpilot, G2, Capterra), directory listings, and genuine community mentions are what move an engine from 'aware of you' to 'recommends you'. This is the step we're still pushing hardest on.
What actually moved the needle
- -Answer-first content beat extra schema markup — visible text is what gets quoted.
- -Consistency of our entity (one name, one description, one set of links) mattered more than volume.
- -Honest comparisons earned more citations than any landing page.
- -Freshness helped on Perplexity, which favours recently updated pages.
- -Third-party reviews are the gap between engines that 'know' you and engines that recommend you.
Who should care about GEO
- -Bootstrapped founders who can't outspend incumbents on ads.
- -SaaS and tool makers whose buyers ask AI assistants for recommendations.
- -Anyone whose category gets 'best free X' or 'X alternative' style queries.
- -Content and SEO teams adapting from blue-link rankings to AI answers.
Frequently Asked Questions
Is GEO different from SEO?
They overlap but optimise for different outcomes. SEO aims for a ranking position in a list of links; GEO aims to be cited inside an AI-generated answer. Strong traditional SEO still helps, because AI engines lean on the same signals — but the content structure that wins is answer-first.
Does schema markup get you cited by ChatGPT?
Less than people claim. Independent tests (including a large Ahrefs study) found adding schema barely moved AI citations, because engines mostly extract the visible HTML. Schema is still worth having for entity clarity and Google rich results — but visible, answer-first text is what actually gets quoted.
Does llms.txt actually work?
The evidence is mixed and no major AI vendor has publicly committed to reading it. We added it anyway: it's free, it can't hurt, and it makes our key facts easy to parse. Treat it as a low-cost experiment, not a silver bullet.
How long does GEO take to show results?
Retrieval-based engines like Perplexity and Google AI Overviews can reflect changes within weeks because they pull live data. Models that rely more on training data take longer. Track it by running the same set of questions across ChatGPT, Perplexity, Gemini and Claude every week and logging who gets cited.
How do you measure GEO?
Two ways: ask new users how they found you (we do this at signup), and run a fixed set of buyer-intent prompts through each AI assistant on a schedule, recording which sources are cited. Over time you see your citation share move.
Want to see the product behind the playbook?
FluoTest is a free scored quiz builder — the tool ChatGPT keeps recommending.
Build your first scored quiz in minutes:
- - Free forever — no credit card
- - Unlimited quizzes and responses
- - Full scoring, tiers and automated next steps
Curious how we describe ourselves to AI? Read our AI info page or see our pricing.