How to choose a GEO provider? What to demand before you pay

Most offers fall into one of two genres.

One is the review. You typically receive a checklist, a presentation and recommendations built on expert opinion. The correctness of the claims is often hard to verify independently.

The other genre is measurement. Its basis is a pre-defined question set, repeated querying and results separated per engine. The process can be re-run, so the result can be held to account.

Your first question should be:

"What will you measure, and can I repeat the measurement?"

1. Real questions

The measurement should be based on questions your buyers actually ask.

An artificially assembled prompt list does not necessarily show how real prospects search and decide.

2. Per-engine breakdown

ChatGPT, Google AI Mode, Gemini and Perplexity are separate surfaces. Their results can differ too.

An aggregate "AI visibility score" easily hides where you actually stand on each engine.

3. Repetition and ranges

Generated answers can vary. No reliable conclusion can be drawn from a single query.

Ask how many times they run the same question, and whether the result is reported as a range, with a margin of error.

4. Raw data

Alongside the report, ask for the full prompt list and the raw answers as well.

Without the underlying data, the result cannot be verified.

5. A follow-up measurement with the same method

The value of the fixes can be proven if the same measurement is later re-run with an unchanged method.

A follow-up built on a different, more favourable question set is not suitable for comparing results.

6. Finished work

A diagnosis is relatively easy to hand over as a PDF. The real value of the service shows in who carries out the necessary changes.

Ask who writes the schema code and the answer-ready copy, and who inserts them into the right surfaces.

7. A fixed price after scoping

A "from" price is acceptable if the provider defines precisely what the base package covers. The price of further work should be given as a fixed amount after scoping.

Open-ended hourly billing is less predictable.

8. A check of access and rendering

The provider must examine whether robots.txt and other technical settings allow AI crawlers in. They must also verify that the essential content is rendered server-side.

If these foundations are missing, the engines may not reach part of the content at all.

9. Unambiguous company facts

The company name, the activity, the prices and the contact details must be clearly identifiable.

Content production built on contradictory or uncertain information gives GEO no stable foundation.

"Guaranteed number-one spot in ChatGPT"

No fixed position can be guaranteed in a generated answer.

Whoever makes such a promise is committing to a result they cannot fully control or reliably measure.

"Secret AI registration" or a "special AI schema"

No such general, secret requirement exists.

Google publicly states that appearing in AI surfaces requires no special markup that exists solely for that purpose.

Success fees without a provable causal link

A success fee makes sense if the provider can show which intervention contributed to which change in the results.

If that cannot be proven, the fee can easily be built on changes the provider had no demonstrable part in.

A score without a methodology

A score is only interpretable if you know exactly what data it is built from and how it is calculated.

Without a methodological description, the number is primarily a marketing device, not a verifiable measurement result.

On the Hungarian market, one-off audits typically start around or above a hundred thousand forints. Monthly fees for ongoing GEO services range roughly between 150 and 400 thousand forints.

On the EU market, one-off audits typically scatter in a band between one and three thousand euros.

The price differences are explained in large part by the genre of the service. It matters whether you get a one-off review or a repeatable measurement — and whether the offer includes only advice, or the actual execution of the changes as well.

Price alone tells you little. The accountability of the service tells you more.

An honest provider speaks plainly about the limits. They tell you what they can measure, and where uncertainty remains.

They can vouch for the quality of their own work. But they cannot fully control the final decision of the answer engines.

Be cautious with a provider for whom everything is solvable and every result can be guaranteed. In this genre, "we cannot guarantee that" is a basic professional requirement.

Before you sign the contract, ask the following questions:

• What do you measure, and what question set do you use?
• How many times do you repeat the queries, and on which engines?
• Do you hand over the prompts and the raw answers?
• What exactly does the base price cover, and what counts as separately billed work?
• Who carries out the recommended changes?
• When and with what method do you re-measure?

If you get a clear, verifiable answer to all six questions, chances are you have found the right provider — whoever they may be.