GEO · Generative Engine Optimization
First we measure whether AI mentions you. Then we talk about money.
GEO is not keywords in a blog post. It is four layers of work: let the models read you, make passages liftable, keep your brand entity consistent, and get cited somewhere other than your own site. Which layers you are missing decides whether this is three weeks or three months — which is why this page has no price list.
Want a look first: the free AI visibility check
Why there is no price list
The same sentence, "do GEO for us", can mean three times the work depending on what is already there. It comes down to the foundations, not the industry or the budget:
| Foundation | Missing it looks like | Having it looks like |
|---|---|---|
| Machine-readable | AI crawlers blocked, content only appears after JS, robots and noindex fighting each other | Server-rendered, crawlers allowed, schema and canonical correct |
| Liftable passages | Long prose, the answer buried in paragraph six, numbers with no unit or date | Definition sentence first, spec tables, FAQ headings that are whole questions |
| Entity consistency | Three spellings of the brand, sameAs pointing at dead links, profiles that disagree | Organization schema and external profiles that name each other |
| Third-party mentions | Nothing outside your own domain mentions you | Comparisons, roundups and forums a model can cite |
The baseline ends in a list: what each of the four layers is missing and what to fix first. That list is what we quote from, and it is yours to take elsewhere. We do not lock you in.
Step one: the baseline
Done before we talk about a retainer. You get numbers and a list, not a pitch deck.
- Question set
- 15–20 questions your buyers actually ask ("best X for Y"), with your brand name kept out so the test cannot answer itself
- Measurement
- Official APIs for ChatGPT, Gemini and Perplexity, each question repeated several times, reported as a mention rate with a 95% confidence interval
- Read-back
- Per question: which brands the answer named and which URLs it cited — that competitor list and those sources are the next target
- Technical scan
- The four foundations checked one by one: crawlers, render, schema, entity, external mentions
- Change list
- Ranked by payoff, marked with what we can do and what needs your engineers or your lawyers
Then: the weekly loop
What the retainer actually buys, all of it pointed at the numbers.
- Rewrite passages
- Start with pages that were consulted but not cited — the model read you and still left you out, which is the cheapest thing to fix
- Fill gaps
- Add citable passages for the questions you lose, not more word count
- Earn mentions
- Comparisons, roundups, forums — what a model cites is usually not your own site
- Re-measure
- Rerun the question set weekly and read the trend with its interval, not the week-to-week wobble
- Monthly report
- What changed, which questions moved, what comes first next month
What we do not promise
- No guarantee that ChatGPT will recommend you. Nobody can guarantee that, and anyone who says it is selling you something else.
- No date. Moving one question from unmentioned to consistently named runs in quarters, not weeks, in what we have seen.
- A good week is not a result. Ask the same question three times in one day and the answer changes, so we report intervals, not points.
- Engines change. A model version or an API behaviour change moves the numbers; when we change a model we log it as a change of method, not a detail.
How the measurement works
Vendor APIs only. We do not drive the consumer chat apps and we do not route through multi-model gateways, because gateways strip the citation metadata half of these metrics depend on. Model identifiers live in one configuration block, and a model change counts as a change of method. The full methodology — endpoints, pinned models, metric definitions and limits — lives on one page, and we do not restate it here:
FAQ
- Why a baseline first? Why not start fixing?
- Without a baseline there is no way to show the work did anything. GEO results are probabilistic — the same question asked repeatedly gives different answers — so with no starting number, three months later you would just have our word for it. The baseline is also what the quote is based on.
- How is this different from SEO? Do I need both?
- SEO ranks a page in search results; GEO gets the generated answer to name you. They share a content foundation and are measured differently: SEO tracks rank and clicks, GEO tracks mention rate and citation rate. The technical base — server rendering, schema, speed — feeds both, so whichever you start with helps the other.
- Can I do this myself?
- Yes. The measurement is available as a subscription you can run yourself, and the method is published. If you have the time, start there — this page is for the half of the job that is time, not knowledge.
- Will you use us as a case study?
- Not without written consent. The baseline and the weekly numbers are yours.
- How do we start?
- Message us on Telegram with the domain and the buying situation you want to be recommended in. We will say up front whether it is worth doing, including when it is not.
Start with the baseline
Send the domain and the buying situation you want to show up in. We look at the foundations first, then talk scope and price.
Want a look first: the free AI visibility check. Want to watch the number yourself: the monitoring subscription