GEO Visibility Index DACH & GCC · Baseline

Which agencies do AI assistants recommend? 525 answers, measured and published.

We asked ChatGPT, Perplexity and Gemini 35 buying questions about GEO and AI search visibility, five times each. Here are the results, the method and the full dataset. Including our own result: zero mentions.

525

measurements

596

firms named

2.2 %

share of voice of the most-named provider

37 of 105

rows with a provider named in all five runs

0

mentions of Grovia Digital

What we measured

Prompts
35 frozen prompts, 15 in German and 20 in English, in five categories: head terms, agency selection, alternatives, regional and methodology questions.
Platforms
ChatGPT, Perplexity and Gemini, each in a new empty chat with web search on. The model versions are recorded per measurement.
Runs
Every prompt five times on every platform: 35 × 3 × 5 = 525 measurements.
Conditions
Measured 2 to 9 September 2026. VPN exit Frankfurt, memory and personalisation off, except where a limitation declares otherwise.
Counting
A provider counts once per answer. Spelling variants are merged. Names of individual persons are removed.

What the answers show

1

There is no market leader.

Across 525 answers the assistants named a provider 2,548 times, spread over 596 firms. The most-named provider reaches a share of voice of 2.2 %. 255 providers were named exactly once.

2

The type of question decides whether names appear at all.

Questions asking for alternatives produced 9.3 providers per answer on average. All 90 answers to methodology questions, such as how to measure AI visibility, named none. There, the only way to appear is to be cited as a source.

3

The three platforms answer very differently.

Perplexity named 6.9 providers per answer, ChatGPT 4.7 and Gemini 3.0. 79 of 175 Gemini answers named no provider at all, against 60 for ChatGPT and 46 for Perplexity.

4

Three runs make answers look more stable than they are.

After three runs, 49 of 105 prompt and platform rows had at least one provider named every time. After five runs, 37 remained. A quarter of the apparent stability disappears with more runs.

5

Gulf answers are more stable than German ones.

In the Gulf block, 11 of 18 rows had a stable core of providers. In the German-language block it was 9 of 45.

6

Most sources are the providers’ own websites.

3,335 cited sources from 1,044 domains: 51.5 % company websites, 19.6 % comparison lists, 12.3 % trade media, 5.4 % press release portals, 3.1 % agency directories. In about 49 % of provider mentions, the provider’s own website was cited in the same answer (an approximation, method in the changelog).

Our own result

Grovia Digital was named in 0 of 525 answers and cited as a source once (Gemini, run 1 of one English prompt), a citation rate of 0.19 %. This is our baseline. We re-measure every two weeks and repeat the full measurement after three and six months, under the same conditions.

What this measurement does not show

  • It is a snapshot of one week in September 2026, not a trend.
  • Five runs per prompt and platform are few. Treat single rows as indications, not proof.
  • All measurements ran via a Frankfurt VPN exit. Answers for other locations can differ.
  • How often a provider is named says nothing about the quality of its work.
  • 29 limitations are declared in full in the dataset, including 35 answers whose sources were not recorded.

Diary: every pulse since the baseline

Every two weeks we repeat a fixed subset of the prompts, three runs each, under the same conditions. Each pulse gets one dated entry here, also when nothing changed.

DateMeasurementMentionsStatus
5 Nov 2026Pulse P4, 90 measurements1Perplexity named Grovia Digital first on a buyer question, citing an answer page that had gone live seven days earlier
22 Oct 2026Pulse P3, 114 measurements6First ChatGPT mentions: ChatGPT named Grovia Digital on buyer questions about agencies in the UAE in 6 of 12 measurements
8 Oct 2026Pulse P2, 90 measurements2First mentions: Perplexity named Grovia Digital on one buyer question (2 of 3 runs)
24 Sep 2026Pulse P1, 114 measurements0No mention yet, first measures live
10 Sep 2026Baseline, 525 measurements0Starting point documented, first measures live

Download the data

Five CSV files (UTF-8, semicolon-separated), a README with the counting rules and a changelog. Free to use under CC BY 4.0.

All files as ZIP
  • Prompts

    35 frozen prompts with ID, language, category and exact wording

    CSV
  • Measurements

    525 rows: date, platform, model, run, location, providers and tools named

    CSV
  • Sources

    3,335 cited sources with domain, URL, source type and position

    CSV
  • Providers

    Every provider named, in alphabetical order, with the number of measurements

    CSV
  • Limitations

    All declared limitations of the measurement, in full

    CSV
  • README

    Markdown
  • Changelog

    Markdown

Pulse data

Pulse files as ZIP
  • Pulse P1 · Measurements

    114 measurements on 24 Sep 2026 (German and Gulf prompts), 3 runs

    CSV
  • Pulse P1 · Sources

    525 cited sources

    CSV
  • Pulse P2 · Measurements

    90 measurements on 8 Oct 2026, 3 runs

    CSV
  • Pulse P2 · Sources

    486 cited sources (Perplexity)

    CSV
  • Pulse P3 · Measurements

    114 measurements on 22 Oct 2026 (German, international and UAE prompts), 3 runs

    CSV
  • Pulse P3 · Sources

    763 cited sources (Perplexity)

    CSV
  • Pulse P4 · Measurements

    90 measurements on 5 Nov 2026 (German and international prompts), 3 runs

    CSV
  • Pulse P4 · Sources

    548 cited sources (Perplexity)

    CSV

Also on: GitHub · Zenodo · Kaggle

How to cite

Grovia Digital (Grovia Management Consultancy L.L.C), Knodel, A., Knodel, K. (2026). AI Visibility Index. Zenodo. https://doi.org/10.5281/zenodo.23385485

Licence: CC BY 4.0: creativecommons.org/licenses/by/4.0

Frequently Asked Questions

Grovia Digital, a GEO and performance marketing agency based in Business Bay, Dubai. We measured ourselves as one of the agencies in this market, and published our own result.
Yes, under CC BY 4.0. You can share and adapt it, also commercially, as long as you credit Grovia Digital and link to this page.
Because how often a provider is named is not a statement about its quality. The provider list in the dataset is alphabetical on purpose.
Yes. We re-measure every two weeks with a subset of prompts and repeat the full measurement after three and six months. Each version gets its own date and changelog entry.

Want this measured for your brand?

We run the same measurement for clients, with prompts built from how their buyers search.

AI Visibility & GEO