Generative engine optimization (GEO): how to get your business named by ChatGPT, Claude and Perplexity

Key takeaways

• GEO means generative engine optimization. It is about being named and cited inside AI answers, not ranked in a list of links.
• It doesn't replace SEO. Assistants with web search start from search results, and Google says its AI features use the same fundamentals as regular Search.
• The pages an assistant reads decide who it names, and many of those pages aren't yours: review sites, directories and local lists matter a lot.
• What helps is ordinary good practice done carefully: a direct answer early, facts with dates and sources, consistent business details everywhere, and structured data.
• Measure it every week across several assistants. One check in one assistant is a snapshot, not a result.

What is generative engine optimization?

A generative engine is any tool that writes an answer instead of showing a list of links: ChatGPT with search on, Perplexity, Claude with web search, Microsoft Copilot, and Google's AI Overviews and AI Mode. Generative engine optimization is the practice of making your content and your business the material those answers are built from.

The term comes from a research paper by Pranjal Aggarwal and colleagues, first posted in November 2023 and published at the KDD 2024 conference. They built a benchmark of real queries, changed the source pages in controlled ways, and measured how often the generative engine used each page. Some changes raised a page's visibility in the answers by up to 40%. Adding citations to sources, quotations from relevant experts and concrete statistics helped. Stuffing in keywords did not. They also found the effect varied by subject, which is a good reason to measure your own case rather than copy someone else's checklist.

For a local business the practical version is simpler. When a customer asks an assistant "who's a good dentist in Kitsilano?", the assistant names two or three businesses. GEO is everything that makes yours one of them, honestly.

GEO is not geotargeting

"GEO" causes confusion because the same three letters have long meant geography in marketing: geotargeted ads, geofencing, or local SEO for a city. Generative engine optimization has nothing to do with location by itself. It shares a lot with local SEO for a local business, because the questions customers ask include a city, but the acronym refers to generative AI, not to maps.

How does GEO work?

It helps to follow what an assistant does with a question, because every GEO step maps to one stage of it.

1. The question is split into searches. Google calls this query fan-out: AI Overviews and AI Mode may run several related searches across subtopics and data sources to build one response. ChatGPT and Claude show the searches they ran, and they are often more specific than the customer's question ("Invisalign price Kitsilano", "dentist open Saturday West 4th").
2. The searches return pages. For Google that is its own index; other assistants use their own crawlers and search partners. A page that isn't indexed, or that blocks the assistant's crawler, is out at this stage.
3. The assistant reads a handful of those pages and picks the passages that answer the question best. A clear, self-contained answer near the top of a page is easy to use; a sales paragraph that never states a fact is not.
4. It writes the answer and names the businesses those passages describe, often with links to the pages it used as sources.

GEO vs SEO vs AEO

SEO gets your pages ranked in search results. AEO, answer engine optimization, gets your content used as the direct answer: a featured snippet or a spoken answer from a voice assistant. GEO gets your business named in an AI-written answer. Most of the work overlaps, because all three reward a clear, trustworthy answer to a real question.

The differences are in what you measure and where you compete. SEO measures positions and clicks. AEO measures whether you hold the snippet. GEO measures how often assistants name you and cite your site, and the competition is for two or three names in an answer rather than ten links on a page. We compare them side by side in our guide to SEO vs GEO vs AEO.

In practice that means one to-do list, not three. A page that is indexed, answers the question in its first two sentences, states its facts with dates and sources, and carries FAQ markup serves all three at once. The only work that is specific to GEO is checking what the assistants actually say, and getting onto the outside sites they lean on.

Is SEO dead now that AI answers questions?

No, and the reason is mechanical rather than a matter of opinion. Assistants with web search start from search results, so a page that search engines can't find or index won't be read. Google says plainly that there are no additional requirements or special optimizations needed to appear in AI Overviews or AI Mode: a page must be indexed and eligible to show in Search with a snippet. What has changed is how much a single click is worth and where the customer meets you. More decisions now happen inside the answer, so being named there matters as much as ranking.

You will also see the same idea called AI SEO, LLM SEO, AI search optimization or AEO. The labels differ; the work is the same. If anyone tells you SEO is dead and only their new acronym matters, ask them how the assistant finds pages in the first place.

How assistants decide which businesses to name

When an assistant has web search on, the businesses it names are mostly the ones described clearly by the pages it reads. OpenAI, Perplexity and Anthropic each run their own crawler for this, and a website that blocks them can't be read. Google's AI features work from Google's own index.

Two things follow. First, the pages the assistant reads matter more than any single ranking signal. Second, those pages aren't only yours. In our own measurement, the assistants built answers about real estate agents largely from third-party review and directory profiles, and the three assistants often relied on different sources for the same question.

What we measured (our own test accounts, October 2026)

In week 42 of 2026 we tracked our two founder test accounts: a real estate practice in six Metro Vancouver cities, and RealBooster itself in six cities across Canada and the US. That was 450 answers from three assistants, two runs per question. These are our own accounts, so treat the numbers as one honest example, not an industry average.

• The real estate practice was named in 0 of 180 answers: the honest starting point for most local businesses.
• Two agent-review directories were cited in 54 and 41 of those 180 answers. Third-party profiles carry a lot of weight.
• For some questions the three assistants agreed on 0% of the sources they cited. Each one built its answer from different pages, so one check in one assistant is a snapshot, not a verdict.

How to appear in AI answers: the white-hat checklist

1. Make sure assistants can read your site: real text in the page (not only loaded by JavaScript), no robots.txt block on OAI-SearchBot, PerplexityBot or Claude's crawlers, and every important page indexed in Google and Bing.
2. Answer the real questions. Write one page per buyer question ("emergency plumber in Burnaby", "Invisalign cost in Kitsilano"), with the direct answer in the first two sentences and your business name and area in it.
3. Cover the searches behind the question. Because assistants fan a question out into several searches, a page that also answers the obvious follow-ups (price, hours, area, how to book) gets picked for more of them.
4. Use facts an assistant can check: licences, years in business, service areas, prices or price ranges, each with a date and, where possible, a source. This is the change the GEO research found most useful.
5. Keep your name, address, phone and category identical on your site, Google Business Profile and the main directories for your trade.
6. Get onto the sites assistants cite for your questions: the review platforms, directories and local lists that show up in their answers.
7. Mark up your pages with structured data (LocalBusiness or your trade's type, FAQPage) so machines read your facts without guessing.
8. Earn real reviews from real customers, and reply to them. Never buy, gate or incentivize reviews.

A worked example: one question, start to finish

Take a family dental clinic in Kitsilano and the question "Who does Invisalign in Kitsilano, and what does it cost?". This is an illustration of the process, not a measurement.

First, find out what the assistants do with it. Ask ChatGPT, Claude and Perplexity with web search on, twice each, and write down who they name, which pages they link, and, where the assistant shows them, the searches it ran. Typical follow-up searches for a question like this are about price ranges, whether a consultation is free, how long treatment takes, and whether the clinic takes new patients.

Next, compare those cited pages with the clinic's own site. Often the clinic has a single "Services" page that lists Invisalign in one line, with no price range, no consultation details and no mention of the neighbourhood. The assistant has nothing to quote, so it quotes someone else.

The fix is one answer-first page: a two-sentence direct answer naming the clinic, the neighbourhood and a dated price range; a short section on how treatment works there; the facts behind the follow-up searches; the dentist's credentials with a link to the provincial college register; FAQ markup; and consistent details on the Google Business Profile and the two directories the assistants cited. Then the clinic waits, measures every week, and changes nothing else for a few weeks so the effect can be read.

How to measure GEO

Google doesn't report AI Overview or AI Mode traffic separately: Search Console includes it in your overall Search numbers, under the Web search type. And assistants give different answers on different runs. So GEO has to be measured directly, by asking the questions and recording the answers.

• Pick the 10 to 30 questions your customers actually ask, with your city in them.
• Ask each assistant with web search on, at least twice, and repeat every week.
• Record four things: whether you were named (mention rate), whether your site was linked (citation rate), who was named instead (share of voice), and which sites were cited (your source list).
• Weight the questions by real demand where it's measured, so a question 900 people ask counts for more than one 10 people ask.
• Change one thing at a time and keep a dated log, so you can tell what moved the numbers and what was noise.

Benefits and challenges

The benefit is reach at the moment of decision. A customer who asks an assistant for a recommendation is close to choosing, and the short list it gives is far shorter than a page of search results. The same work also helps ordinary SEO and makes your website clearer for people.

The challenges are real. Answers vary between runs and between assistants, so results are noisy. You don't control the third-party sites assistants read. Some assistants show sources, some don't. And it takes time: in our experience first citations usually appear 4 to 8 weeks after the right pages and profiles exist, and Google ranking movement takes 3 to 6 months. It varies by trade, city and competition, and it should be measured, not promised.

What doesn't work

• Hidden text, keyword stuffing or pages written for bots instead of people. The GEO research found keyword stuffing didn't help.
• Fake reviews, paid links or link swaps. They break Google's policies and get the business penalized.
• Claims you can't back up. An assistant that finds a contradiction elsewhere tends to trust the other source.
• Anyone who guarantees you a spot in ChatGPT. Nobody controls the answers, so a guarantee is a red flag.

Can you do GEO yourself?

Yes, for the basics. If you can edit your website and your Google Business Profile, you can do the checklist above: answer-first pages, consistent details, real reviews, structured data with a plugin or your developer's help. The part that's hard to do by hand is the measuring: dozens of questions, several assistants, two runs each, every week, with the sources recorded. That's the part most owners skip, and without it you can't tell whether anything worked.

How RealBooster does it

RealBooster measures your AI visibility every week across ChatGPT, Claude and Perplexity, plus Google's AI Overview, for the questions your customers ask in your city. Then it does the work: answer-first pages built on the searches the assistants actually ran, consistent business facts, and the profiles that feed the answers. We sell a limited number of seats per trade in each city, so we never work for your direct competitor in your area.