GEO (generative engine optimization) is the work of getting your brand named, and described correctly, when people ask ChatGPT, Perplexity or Gemini what to buy. It is not a new trick. It is SEO plus being talked about on the right websites, measured differently. This page shows you how AI picks brands, runs a 20-minute check on yours, and tells you the first thing to fix.
Where to start
Ask AI about your category the way a buyer would
Here is how an AI engine picks brands, in three moves. It turns the question into web searches, reads a small set of pages from the results, then names the brands those pages name. OpenAI says ChatGPT rewrites your question into shorter search queries and can use your rough location and saved memories. So every page it reads carries real weight.
That is why GEO looks like this next to classic SEO:
| Classic SEO | GEO | |
|---|---|---|
| The win | A high spot in a list of links | Your brand named in the answer, with the right price and use |
| Bot that must reach you | Googlebot | Also OAI-SearchBot (ChatGPT search) and PerplexityBot |
| What moves you | Your pages and backlinks | Your pages plus what other sites say about you: lists, reviews, Reddit, YouTube |
| How you check | One search, a fairly stable rank | Many answers, because each one names different brands |
Now test it. Fill in your details and you get three sets of prompts. Run the discovery prompts first: they leave your brand name out, like a real buyer would. A question that names you will always mention you.
Test-prompt pack builder Example brand
Runs in your browser. Nothing is sent anywhere. Your answers are saved in this browser for next month's check.
A. Discovery prompts: does AI name you?
B. Brand prompts: does AI describe you correctly?
C. Source prompts: ask right after a discovery answer
Turn your count into an honest visibility score
AI answers change every time. The same question can name different brands on each run, so one check can mislead you. The fix is to count answers and read the result as a range, the way pollsters report a margin of error. Enter your Step 1 tallies below.
Visibility share calculator Example numbers
Runs in your browser. Uses the Wilson score interval, a standard formula for percentages from small samples.
Find your first 3 fixes
Most brands that AI skips fail at one of four things: AI cannot read them, their facts are not quotable, other sites do not mention them, or nobody measures. Answer six questions honestly. You get a readiness score and your first three fixes in order.
GEO readiness diagnostic Example answers
Runs in your browser. Tap one answer per question. Your answers are saved here.
Fix: one description, used everywhere
AI can build its picture of you from several pages. If your site, marketplace listings and profiles disagree, the answer can repeat the wrong version. This sentence is built from your Step 1 answers. Edit it once, then paste it on your homepage, About page, marketplace seller pages, LinkedIn and Google Business Profile.
Fix: write product facts an AI can quote
The second line answers a price, a size, a skin type and a "who should not buy" question at once. Put it in plain page text, not only in an image.
Fix: ask to be added to a best-of list AI already cites
Questions
What does GEO stand for?
Generative engine optimization: making your brand show up, and show up correctly, in answers written by AI engines such as ChatGPT, Perplexity, Gemini and Google AI Overviews. AEO (answer engine optimization) and "LLM SEO" mean the same job.
Is GEO replacing SEO?
No. AI engines with web search read pages that search engines return, so the SEO basics still count: crawlable pages, clear answers, a good reputation. GEO adds two things: being mentioned on the sites each engine reads, and measuring across many answers instead of one rank.
Do I need an llms.txt file?
It is harmless but not a priority. Ahrefs checked 137,210 domains and found 97% of llms.txt files got zero requests in May 2026. Spend the time on Step 3 fixes first.
How long until AI starts recommending me?
No engine publishes a timeline. Sources also turn over fast: SISTRIX found ChatGPT Search replaced 74% of its cited domains each week over 17 weeks. Plan for a steady flow of new mentions, and measure monthly.
Why does ChatGPT name me but Perplexity does not?
Each engine runs its own search and cites different sites, so check all of them. Guide 01 covers this gap engine by engine, with a robots.txt checker.
Data: Ahrefs llms.txt study, May 2026; SISTRIX citation drift, Dec 2025 to Apr 2026; OpenAI help and crawler documentation. Free guide by SEO to GEO HQ.