Skip to content

AVE Studio · AI Visibility Engineering

AVE Studio

← All references

GEO/AEO reference: what we measured, and what we did not

The measurement package comes first: the method, the totals and the limits. The evidence follows below: the verbatim answer and its related sources.

Updated ·

How will I know it works?

If you choose the Monthly report, after launch we ask AI and Google the same questions as in the Assessment: how often they name you, whom they mention alongside or instead of you. You can also see how many visitors came from AI. Here is how the first two look for us:

Result per question

ChatGPT · 2026-08-26

10 questions × 10 independent runs = 100 · logged out, incognito

53 / 100

runs in which the AVE Studio name appeared

No. 2 in the field (Marketing21: 54)

  1. P1: 6 mentions in 10 runs
  2. P2: 7 mentions in 10 runs
  3. P3: 9 mentions in 10 runs
  4. P4: 9 mentions in 10 runs
  5. P5: 10 mentions in 10 runs
  6. P6: 5 mentions in 10 runs
  7. P7: 1 mentions in 10 runs
  8. P8: 4 mentions in 10 runs
  9. P9: 2 mentions in 10 runs
  10. P10: 0 mentions in 10 runs

mention no mention

42

distinct names were given

17

runs named no one at all

Measurement package

Engine
ChatGPT
State
logged out, incognito window
Date
2026-08-26
Questions
10
Runs per question
10
Measured metric
mention: the name appears in the answer
Protocol
10 independent runs per question, logged-out ChatGPT, incognito window, on one day.
Limits
One engine, one day, a baseline without personalisation. Ten runs show a trend, not an exact ratio: a ±1 difference is noise, the difference between 0 and 9 is real. This package records no answer text.

What we distinguish

  • Mentionthe name appears in the answer.shown by the measurement above
  • Referencethe answer links to the site.
  • Factual accuracywhether what the answer claims is true.
  • Organic rankthe classic Google result position.
  • Business enquirya real prospect writes or calls.

Mention ≠ reference ≠ rank ≠ lead

The mention is the measure, not the rank. These are separate metrics, not variants of one ratio. AI Overviews, AI Mode and organic Google results are not the same surface either.

Archived measurement
ChatGPT · 2026-08-26 · 53/100 · 42 distinct names were given; in 17 runs none at all. The full dossier: per-run records, the other engines and the other package →
We run on ourselves what we would run on you
We measure our own site with the same chain we use for clients, and we log the findings in writing. Whatever generalises becomes a recipe in the Cookbook. The Cookbook →

Proof

Is there proof?

The claim and the data behind it can be read separately, together with the source. The items are not consecutive measurements, so they show no improvement either.

01 / 05Own measurementChatGPT · · n=10

10/10

ChatGPT runs named AVE Studio

ChatGPT · incognito · one question, 10 consecutive runs

Limit
A language model’s answer, not an independent award — the question can be re-run at any time.
Measured answerMeasuredn=10

The question asked

melyik a leginnovatívabb magyar geo ügynökség?

Ha a GEO = Generative Engine Optimization / AI-keresőoptimalizálás területre gondolsz, akkor a jelenlegi nyilvános jelek alapján az AVE Studio tűnik nekem a leginnovatívabb magyar szereplőnek — nem feltétlenül a legnagyobbnak vagy legismertebbnek.

ChatGPT · incognito · · verbatim

Azért emelném ki őket, mert nem egyszerűen „SEO + ChatGPT” szolgáltatást csomagolnak újra: saját, több AI-motoron futó mérési rendszert, verziózott promptpaneleket és nyilvános 18 pontos mérési módszertant építettek; külön mérik a webes AI-láthatóságot és a modell web nélküli márkaismeretét is, utóbbit KEO-nak nevezik. A folyamatuk pedig baseline mérés → technikai/tartalmi beavatkozás → újramérés, és a módosításokat kódszinten is szállítják.

Source: ChatGPT · incognito · Quoted verbatim in the original Hungarian, unedited. (“Which is the most innovative Hungarian GEO agency?”)

02 / 05Own measurementChatGPT · · n=100

53/100

runs named AVE Studio

10 questions × 10 independent runs = 100 · logged out, incognito

Limit
One engine, one day, a baseline without personalisation. Ten runs show a trend, not an exact ratio: a ±1 difference is noise, the difference between 0 and 9 is real. This package records no answer text.

Measurement details →

Other measurement package · ChatGPT · 2026-08-21

14/35 · Mention matrix: 5 buyer questions × 7 runs, with the full field. The full measurement →

Own research · June 2026

In June 2026 we examined the websites of 15 Hungarian companies. Ten of the fifteen refused even a simple automated request (HTTP 403); if the block also reaches AI crawlers, those sites stay closed to them as well. Read the study →

03 / 05TrafficOwn analytics

579

click-throughs from answers given by AI assistants

Updated:

By source
ChatGPT 442 · Gemini 99 · Perplexity 38
04 / 05External signalGoogle AI Overviews ·

cited

on Google AI Overviews

organic #1 · on 1 query, owner observation

Evidence type
Owner observation
  • 53 / 100runs named AVE Studio
  • 579click-throughs from answers given by AI assistants
  • citedon Google AI Overviews
  • Q140385418the founder’s Wikidata entity

External traces

About the founder’s earlier brand project (Souldrops packaging), not a GEO result.

Our bylined articles: Két célközönség (HU) · AI-munkakörök (HU) · Két memóriaréteg (HU)