When a buyer asks ChatGPT, Gemini or Perplexity to recommend a company like yours, the answer names a handful. We measure how often it names you and how often it names your competitors, with Seeklayer, the tool we built for this. Then we work on what the assistants read about you. Every month you get a report with a dated score and the actions in order.
Generative engine optimization, also sold as AI search optimization or answer engine optimization, is the work of getting named when someone asks an AI assistant about what you do. Search Console, Bing Webmaster Tools and Analytics already cover part of it. The open question is how many answers name you, and how many name the company you compete with.
On Google your buyer opens tabs and compares. In an assistant they read one answer with a few names in it. Being inside that answer counts: Seer Interactive found that brands cited in an AI Overview earned 120% more clicks per impression than brands left out of that same answer.
The answer names three or four companies. Every time a buyer asks, one of them takes the slot you wanted.
SEO tracks rankings, clicks and impressions. GEO tracks if you get named, how early in the answer, and which words the assistant picks to describe you.
Google’s AI leans heavily on pages that already rank, and the assistants also read your reviews, your listings and the trade sites covering your industry. SEO lays the base. GEO measures what the assistants do with it.
SparkToro and Gumshoe.ai ran the same prompts almost 3,000 times across ChatGPT, Claude and Google and found less than a 1 in 100 chance that two answers name the same set of brands. So we report your rate of appearance as a percentage instead of a yes or a no.
Your site and your market: what you sell, who buys it, and who competes with you for that buyer.
The ones a buyer would type. “Best X near me.” “Who should I hire for Y in Dallas?” With your brand name in the prompt and without it, because those two come back differently.
Each prompt runs several times in each engine, with web search turned on. The number then counts what repeats, not what happened once.
A 0 to 100 score with the formula in the open: 60% mention rate, 25% position in the answer, 15% presence in cited sources. Every run is stored with its date, so next month has something real to compare against.
One number, its trend, and a short list of what to move next.
Seeklayer is ours. We built it at STRADA and we run it on every GEO engagement, which is why the method on this page is the method you get. It sends your prompts to each engine the way your buyer would, keeps what comes back, and works out how often you are named and how often the other companies are. It reads answers from the outside, because no tool sees inside an assistant’s reasoning.
Each prompt runs several times per engine, with web search on. Assistants answer differently from one day to the next, so a batch tells you where you stand and a single run tells you nothing.
Every answer is archived with its date and the sources it cited. That is what makes a comparison possible: you can see what changed, and when it changed.
Being named in the text and having your URL in the source list are two different outcomes. Kevin Indig found that most AI citations link a site without naming the brand. Seeklayer counts them apart, because each one needs a different fix.
Which pages, directories, review sites and trade publications the assistants pull from when they answer about your industry. That list becomes next month’s work order.
Seeklayer is our working tool, so there is no new dashboard for your team to learn. What lands in your inbox is the report, with the measurement done and the actions in order.
ChatGPT, Gemini, Perplexity and Claude. Each engagement starts with the engines your buyer actually opens, and that list goes in writing before the first run.
Every prompt runs with web search turned on, so the answer rests on what is published about you today.
Ahrefs looked at 863,000 searches in 2026 and found that only 38% of the pages cited in AI Overviews sat in the top ten results. The assistant reads around the whole topic, so this gets measured on its own, with prompts written for your business.
Documents that arrive finished, read in one sitting and end in a decision. All five are yours to keep.
Reads in one sitting
An assistant answers with what it finds published about you. We work the levers Google documents and the research supports, because Google treats any attempt to manipulate its AI answers as spam. Every action comes out of the measurement.
Pages that state what you do, where you do it and who for, and that answer what your buyer actually asks before signing.
We check that the search crawlers can read you: OAI-SearchBot for ChatGPT, PerplexityBot, Claude-SearchBot and Googlebot. GPTBot is a training crawler, so that one is a separate call you make with us.
Organization and LocalBusiness in JSON-LD, matching what the page actually says. Google states that its AI features work without special markup, so the markup earns its place on classic SEO grounds.
A complete profile, and your name, address and phone number identical everywhere they appear. Google recommends it for any business that serves customers in person.
The sites the assistants quote when they answer about your category: trade associations, industry press, the review and listing sites people really use. When a competitor is on one and you are not, that is the action.
Who signs your content, what experience backs it, and what customers say in real reviews, asked for without offering anything in return. This is what holds up a good description of you.
What comes up on the first call, answered with what Google and OpenAI publish and what the research shows.
Let’s talk about your projectGEO, AEO and AI search optimization are names for the same work: getting an assistant to name you when someone asks about what you do. Google’s own position is that optimizing for its AI features is still SEO, because the AI needs an indexed, useful page to read. What changes is the scoreboard, since you start tracking mentions inside an answer alongside rankings. ChatGPT, Perplexity and Claude also run crawlers of their own.
Step one is letting ChatGPT read you: OAI-SearchBot allowed in your robots.txt and OpenAI’s IP ranges open at your host. GPTBot is a different crawler and a different decision, since it feeds model training. From there the work is on the information ChatGPT finds about you, measured often enough to tell a change from a coin flip. OpenAI says plainly that nobody can guarantee a position, so what we commit to is method and measurement.
According to Google, the same things as regular search: be indexed, and allow your page to be shown with a snippet. No special file or markup. What Google asks for is useful, original content, plus a well kept Business Profile if you serve customers in person. If you would rather stay out, you control that with nosnippet or from Search Console.
Google ignores llms.txt and describes structured data as optional for its AI features. None of the other engines documents using llms.txt either. Structured data still earns its keep in classic search, so we maintain it for that reason and say which column it belongs in. Your budget goes to the levers that are documented.
Partly, with tools you already own. Search Console has shown impressions from AI Overviews and AI Mode since August 31, 2026, Bing Webmaster Tools shows citations in Copilot, and Analytics picks up sessions arriving from ChatGPT. One question is left over: how many answers name you compared with your competitors. That is the one we measure with Seeklayer, on a published formula of 60% mentions, 25% position and 15% cited sources.
There is usually a checkable reason: search crawlers blocked, pages not indexed, a site that says little about what you do and who for, a thin Business Profile, or few outside sources that mention you at all. The measurement records which sources the assistant cites when it names your competitor, and that list tells you where to start.
It depends on the crawler. The training ones (GPTBot, ClaudeBot, Google-Extended) decide whether your content trains models, and blocking them leaves you visible in ChatGPT search and Claude search. Google-Extended does not touch Google Search, though it does affect the Gemini apps. The search crawlers (OAI-SearchBot, PerplexityBot, Claude-SearchBot) are the ones that make you visible. We go through your robots.txt line by line and you call each one.
You do, and it is written into the contract before we start. That covers the prompt set we wrote for you, the full dated measurement history, every report, and any page, schema or content we produce for your site. Ask for an export at any point and it comes as files you can read without our tool. If the engagement ends, the history and the documents stay with you.
Tell us what you sell and who buys it. We reply within one to two business days and walk you through how we would measure your case: which prompts, which engines, and what lands in your inbox each month.
Let’s talk about your project
