OpenAI
100 Questions
1 mention in 20 answers
2026 public datasetResults published
The 2026 AI Visibility Index
A transparent study of which AI visibility software brands appear in web-grounded answers—and which sources the models cite. Peec AI led the frozen cohort; seven of 25 products appeared at least once.
Registered and collected
Study status
100%
Provider coverage
All 80 planned answers were grounded and score-eligible.
25
Frozen cohort brands
Products with a public first-party AI visibility offering.
20
Frozen discovery questions
Neutral category and buyer questions used for ranking.
80
Grounded discovery answers
Twenty shared questions across four frozen provider models.
Headline results · 100% coverage
Peec AI led a fragmented field
Peec AI appeared in 6 of 80 eligible answers, producing 7.5% discovery visibility. Semrush, Otterly.AI, and Profound each appeared five times. No cohort product appeared in more than one in thirteen answers.
Only seven of the 25 frozen products were mentioned at all. The result is a directional snapshot of this question set and collection window, not a permanent category ranking.
RankBrandVisibilityMentions
017.5%6
Peec AI
026.3%5
Semrush
036.3%5
Otterly.AI
046.3%5
Profound
055.0%4
100 Questions
063.8%3
AthenaHQ
072.5%2
Analyze AI
The complete CSV includes all 25 cohort products, prominence, claimed-domain citations, and zero-mention rows.
Provider differences
The leader changed by model
Anthropic
Semrush
2 mentions in 20 answers
100 Questions + Peec AI
1 mention in 20 answers
xAI
Profound
3 mentions in 20 answers
Cited-source analysis
The answers drew from 210 domains
The leading cited domains were mostly specialist AI visibility sites rather than the largest SEO publishers. A domain is counted at most once per answer.
- 01trakkr.ai10 answers
- 02demandsphere.com8 answers
- 03astiva.ai7 answers
- 04viali.ai7 answers
- 05github.com6 answers
- 06brandpeek.ai5 answers
- 07llmpulse.ai5 answers
- 08100questionsai.com4 answers
Why register it first
The method should not move after the answers arrive
AI answers vary by provider, model, search result, and time. Publishing the protocol first creates an auditable record of what will be asked, what counts as an eligible answer, and how each metric will be calculated.
The published results keep answer coverage beside every headline number. Four exact-string collisions found during manual review were excluded with their question, provider, and rationale preserved in the adjudication file.
Study design
One shared discovery test across four providers
The first edition focuses on software sold for AI visibility, AEO, GEO, or LLM SEO. The frozen 25-product cohort includes only products with a public first-party page offering measurement, monitoring, benchmarking, or optimization of visibility, mentions, citations, or positioning in AI-generated answers. Inclusion is not an endorsement.
Freeze before collection
The cohort, question text, locale, model IDs, matching rules, and scoring version are downloadable before any counted answer collection.
Require sources
An answer is score-eligible only when the provider call succeeds and returns valid web sources. Exclusions remain visible in coverage.
Release the evidence
Publish brand-level metrics, provider splits, cited-domain counts, answer-level evidence where permitted, and machine-readable CSV files.
Frozen question bank
Questions a buyer might actually ask
Discovery prompts do not name a cohort brand. Five diagnostic templates are included in the download for qualitative product checks but will not enter the ranking metrics.
- 01What are the best tools for measuring whether a brand appears in AI-generated answers?
- 02Which tools track brand mentions across ChatGPT, Claude, Gemini, and Grok?
- 03How can a company measure its share of voice in AI search?
- 04What software shows which sources AI assistants cite about a brand?
- 05Which AI visibility tools use repeatable or frozen question sets?
Registered metrics
No composite score without its components
- Discovery visibility
- Eligible discovery answers mentioning a cohort brand divided by eligible discovery answers.
- Prominence
- Mean placement score when a brand is lead, shortlisted, incidental, or absent.
- Provider coverage
- Eligible web-grounded answers divided by planned answers, reported beside every result.
- Citation presence
- Eligible answers citing a cohort brand's claimed domain, kept separate from mentions.
- Source frequency
- The domains cited most often across eligible answers, with repeat citations counted transparently.
Open results release
Every result can be traced back to the collected evidence
Download brand and provider tables, cited-source frequency, full answer text with source URLs, the registered inputs, manual adjudications, and a manifest containing file hashes.
Release checks passed
- • 80 of 80 answers grounded
- • All four model IDs matched
- • 34 positive alias matches reviewed
- • Six result files hashed in a manifest
Limits
What the Index shows—and what it cannot prove
It can show
- Directional inclusion and prominence in one frozen test.
- Differences among provider APIs using required web search.
- Which domains were cited in eligible answers.
- How answer coverage affects the reported metrics.
It cannot prove
- Permanent rankings or every possible buyer question.
- Parity with consumer chat interfaces or personalized sessions.
- That a citation caused a recommendation.
- Statistical representativeness beyond the registered question set.
Study questions
About this first release
What is the 2026 AI Visibility Index?
It is a public study of which AI visibility, AEO, GEO, and LLM SEO software brands appear in web-grounded answers to a frozen set of buyer questions. The protocol is published before collection so the questions and headline metrics cannot be changed to fit the results.
Does this page contain rankings yet?
Yes. The results include all 25 cohort products, provider-level splits, cited-domain frequency, and answer-level evidence. Seven cohort brands appeared at least once across the 80 eligible discovery answers.
Which AI providers did the study test?
The study tested OpenAI, Anthropic, Google, and xAI through their APIs with required web-search grounding. The exact model identifiers, collection timestamps, and result-file hashes are included in the downloadable manifest.
Will the Index represent every consumer AI session?
No. API responses can differ from consumer chat products because prompts, routing, personalization, models, and search behavior can differ. The Index is a time-stamped directional comparison of one controlled test.
Related research
Inspect the benchmark method behind the Index
Review grounding eligibility, metric denominators, evidence retention, and interpretation limits in the full methodology.