Methodology

Engines and exact versions

A study that does not name the model version it ran against is not reproducible, because the answer changes when the model does.

Engine	Model version
Claude	rankxa-warehouse
Gemini	rankxa-warehouse
ChatGPT	rankxa-warehouse

Run window, UTC: 2026-09-04T09:45:00Z to 2026-09-25T04:00:06Z.

Method

The design question for a study like this is not "what do AI assistants think of my company", which is not answerable, but "what do AI assistants say when asked", which is. The method here is a prompt-and-capture design: write realistic buying questions, send them to each engine the same way a prospective customer would, and record whether a given company's name appears in the answer and, if so, in what position.

The run. Over a run window from 2026-09-04 to 2026-09-25 (2 boundary timestamps marking start and end), the study sent 8,188 distinct prompts across 3 engines, each running a single pinned model version (one version per engine, 1 tracked per engine), and collected 536,301 rows of raw output. Of those, 536,301 came back as usable answers, a completion rate that means the 536,301 figure used throughout this study is, for practical purposes, the whole attempt, not a filtered subset.

The prompts. Prompts were split into four types, reflecting different stages of a buying journey: informational ("what is X"), navigational ("where do I find X"), commercial-investigation ("best X for Y"), and transactional ("who should I hire for X"). Each type returned enough answers to stand on its own:

Prompt type	Answers	Appearances	Appearance rate
Commercial-investigation	56,085	21,920	39.1%
Informational	40,245	15,576	38.7%
Navigational	54,573	20,531	37.6%
Transactional	385,398	147,465	38.3%

The four rates sit close together, within a few points of each other, which is itself a finding: whatever drives whether a company gets named, it is not strongly a matter of which stage of the buying journey the question represents.

What this cannot see. A pinned model version per engine means the study measured a snapshot, not a trend. Model providers update their systems without notice, and a re-run in a later month could show different numbers for reasons that have nothing to do with any company's own actions. The study also cannot see why a model chose to name one company over another, only that it did. Mechanism is inferred in later sections, not measured directly.

Which models answered. Each engine was queried through its vendor's developer API, not through the consumer ChatGPT, Claude or Gemini apps, which add account history, personalisation and their own defaults. Almost every answer in this window came from the vendor's small, fast model tier responding from the model's own knowledge, with web search switched off: OpenAI's GPT-4o mini, Anthropic's Claude Haiku and Google's Gemini Flash Lite. In mid September Rankxa moved to newer models with web search switched on. Less than one percent of this edition's questions were answered on the new setup, so this edition remains comparable with July and August. The next edition will be the first on the new setup, and it will say so.

Limitations we volunteer

Written by us, before anyone else found them.

Terms used in this study

Appearance rate

The share of answers to a given prompt type in which a company was named at all, regardless of position.

Visibility

Whether a company was named by a given engine at least once across the study's prompts, reported per engine and as a share of the cohort.

Average mention position

Where in an answer's list of named companies a given mention falls on average, with lower numbers meaning earlier mentions.

Pair agreement

A measure of how often two or more engines, when both naming companies for the same prompt, name the same company, used here to show how often engines converge versus diverge.

Sole mention

A case where only one of the three engines named a given company for a given prompt, while the others did not.

Prompt type

A category describing the intent behind a prompt, such as informational, navigational, transactional, or commercial-investigation, used to see whether the kind of question asked changes which companies get named.

Cohort

The full set of companies tracked in this study, against which visibility and never-appears figures are measured.

Run window

The start and end timestamps, in UTC, over which the study's prompts were sent and answers collected.

Usable answers

The count of engine responses that were successfully returned and parsed for company mentions, excluding any failed or malformed responses.

Never appears

The count and share of cohort companies that were not named by any engine on any prompt during the run window.

References

Sources this study reads against. Every link was fetched and confirmed reachable at publication.

1. GEO: Generative Engine Optimization arXiv / KDD 2024, 2024 Formalizes generative engines as synthesis-based search systems and proposes a 10,000-query benchmark showing visibility gains from citation-style content interventions; the one peer-reviewed source in this space, used to frame the mechanism question our study does not test.
2. AI Visibility Statistics (2026): How Often Brands Appear in ChatGPT, Perplexity, Claude & Gemini Boring Marketing, 2026 Reports an overall AI citation rate and share of brands invisible across all platforms checked, used as the nearest published comparator to our never-appears figure.
3. Cracking the ChatGPT Code: How to Grow Brand Visibility in OpenAI Yext, 2026 Finds that ChatGPT and Gemini draw citations from brand-owned sites versus managed listings at different rates, based on 17.2 million citations; used here as a comparison point for cross-engine disagreement.
4. Search results Anthropic Claude Platform Docs, 2026 Notes that citations are disabled by default and must be explicitly enabled per search result, a configuration detail relevant to cross-engine comparability.
5. Web search tool Anthropic Claude Platform Docs, 2026 Documents the structured citation fields (url, title, cited_text) Claude attaches to web search results, relevant to how position and naming were extracted for Claude in our pipeline.
6. GEO: Generative Engine Optimization - Princeton University Princeton University Collaborative Research Database, 2024 Formal publication record for the GEO paper, confirming KDD 2024 proceedings venue and page range.
7. The Most Cited Domains by LLMs Similarweb, 2026 Finds Wikipedia and Reddit each account for roughly 12-13% of ChatGPT citations in the US, based on nearly 600,000 citation events; cited for comparison of concentration versus our pairwise agreement finding.
8. The Most-Cited Domains in AI: A 3-Month Study Semrush, 2026 Tracks a sharp swing in ChatGPT's Reddit citation share over two months, used to illustrate volatility in single-engine citation behavior over time.
9. AI search engines cite Reddit, YouTube, and LinkedIn most: Study Search Engine Land, 2026 Reports Peec AI's analysis of 30 million sources finding Reddit the most-cited domain across five engines; cited to distinguish source-domain studies from our company-naming study.
10. Gartner Survey Finds 69% of B2B Buyers Turn to Sales Reps to Validate AI-Generated Insights BusinessWire / Gartner, 2026 Survey of 645 B2B buyers finding an average of seven information sources used per purchase and 45% using GenAI primarily to gather vendor information; provides the buyer-behavior rationale for why engine-naming matters commercially.
11. ChatGPT search for Enterprise and Edu OpenAI Help Center, 2026 Confirms citation behavior is consistent for enterprise and edu tiers of ChatGPT search.
12. Searching the web with ChatGPT OpenAI Help Center, 2026 Describes how ChatGPT renders inline citations, used to explain the mechanism behind what our study counts as a 'mention'.
13. Gartner: Two-thirds of B2B buyers prefer rep-free purchasing as AI reshapes sales Digital Commerce 360, 2026 Secondary report on the same Gartner survey, quoting Gartner's Alyssa Cruz on buyers progressing through buying tasks more autonomously.
14. LLM Citation Study: 13,184 Citations Across 4 AI Tools Orbit Media, 2026 Documents a gap between brand mention rate and formal citation rate on Claude and Gemini; used to distinguish 'named' from 'cited' in our own methodology.
15. Top 10 Sources LLMs Cite Most in 2026 Contently, 2026 Evertune's analysis of 200 million prompts finds even the top-cited domain on any platform rarely exceeds five percent of citations, supporting a picture of a long tail rather than concentrated citation.
16. AI Search Visibility Study (2026) Minuttia / Convex Works, 2026 Analyzed 24,067 prompts across five engines and found most recommended brands appeared in only one engine's answer; closest published analogue to our pairwise agreement measurement.
17. Enable and use web search Anthropic Support, 2026 States that when Claude's web search tool is invoked, citations are included by default in the surfaced response.
18. AI Visibility Index 2026 NeuroRank, 2026 Examines 122 brands across four engines, a much smaller cohort than ours, included to note the scale difference in the literature.
19. State of AI Citations 2026: Who AI Engines Cite 5WPR, 2026 Synthesizes over 680 million tracked AI citations and identifies brand search volume as the strongest predictor of citation likelihood; cited for scale comparison and predictor claim.

Data

The complete row-level dataset is published open and ungated under CC BY 4.0. Every number in this study can be recomputed from it.