What does an agency actually offer when its website says “GEO,” “AEO,” or “AI visibility”?
At IndexDock, we studied public agency and consultant offers across 30 European countries. The working database contained 1,051 records, consolidated to one record per domain. Collection took place on 6–8 July 2026.
This was a study of publicly described offers, not a ranking of agencies or an evaluation of their delivery.
We compared the work described, not just the labels
Of the 1,051 records, 632 were classified as active. A relevant offer was confirmed in 625; seven remained unknown.
For each offer, we looked for four explicitly described stages:
- Analysis: an audit or diagnostic of current visibility in AI answers or AI search.
- Strategy: a plan or recommendations derived from analysis.
- Technical implementation: specific work on markup, structured data, rendering, crawler access, llms.txt, or page structure.
- Monitoring: ongoing tracking of visibility in AI answers, or continued management after delivery.
The phrase “technical SEO” alone did not qualify as evidence of technical implementation. Recording llms.txt as an advertised task also does not establish its effectiveness.
Only 195 records described all four stages
Across the 632 active records:
| Stage explicitly described | Records | Share |
|---|---|---|
| Analysis | 382 | 60.4% |
| Strategy | 421 | 66.6% |
| Technical implementation | 235 | 37.2% |
| Monitoring | 280 | 44.3% |
All four stages were described in 195 records — 30.9%.
For 198 records, no stage could be established from the checked public pages. That means unknown, not that those providers perform no such work.
A website can describe its services poorly while its team delivers well. Public presentation and delivery quality are different questions.
Search wording changed the websites we found
A separate component ran 250 general web-search queries across the same 30 countries.
For provider-seeking queries, we compared country-specific wording with Europe-wide wording. The median overlap, measured against the combined set of returned domains, was 11%.
This suggests a practical consideration for anyone designing discovery or monitoring tests: geography and wording can substantially change the source set.
These were web-search results, not AI-answer tests. We did not query ChatGPT, Perplexity, Gemini, or AI Overviews. A returned domain was not counted as an AI citation.
What we would ask before comparing offers
The service label is a starting point. The more useful questions are:
- Which tasks are included?
- Who implements changes?
- Which search or answer-system outputs are being assessed?
- What does monitoring measure?
- What can the evidence actually establish?
The sample was assembled through search and public sources. It was not a census or a statistically representative market estimate, and checking depth varied between countries.
Disclosure: IndexDock sells AI-visibility services in this market. Its own website was excluded from the dataset.
Read the full study, tables, definitions, and limitations on IndexDock.
How do you distinguish a concrete implementation scope from an AEO/GEO service label?
Top comments (0)