Methodology
Everything you need to reproduce, audit, or cite our weekly numbers.
Prompt selection
For each industry × city pair we curate 15–30 prompts a real customer might ask an AI engine — mixing category, comparison, and 'best in [city]' phrasings. Prompts are reviewed by a human, then locked for the quarter so week-over-week deltas are apples-to-apples.
Engines queried
Every prompt is run against four engines: ChatGPT (GPT-4-class), Perplexity (sonar-pro), Google AI Overviews (via a fresh search session), and Gemini. All engines are queried from the same US geography and account tier for consistency.
Run count
Each prompt is asked 4–5 times per engine per week. AI answers are stochastic, so a single response is noise. Averaging over multiple runs surfaces which brands the model actually associates with the prompt, not which one it happened to name once.
Mention extraction
Responses are parsed for named brand mentions. A mention counts if the engine names the brand as an answer or explicit recommendation — passing references ('similar to X') don't count. Mentions are deduplicated per response so a brand named three times in one answer counts once.
Publication cadence
Data refreshes weekly on Mondays at 09:00 ET. Each entry carries its own last-checked timestamp so anyone citing the numbers can point to a specific version. Historical weeks are archived and remain permalinked.
License & attribution
The dataset is free to use under CC BY 4.0. Attribute as: Mostly Marketing AEO Citation Index, https://mostlymktg.com/citation-index. Machine-readable JSON is available for every entry at /api/public/citation-index/[slug].
What we don't claim
AI engines evolve. A brand's rank this week reflects how the current generation of models responds to the current prompt set — not a permanent measure of quality or reputation. We publish deltas so readers can see the direction, not just the level.
How to cite us
The dataset is free to reuse — including commercially — under the Creative Commons Attribution 4.0 International (CC BY 4.0) license. The one requirement is credit with a link back.
Mostly Marketing AEO Citation Index. [Industry, City] — Week [N] [Year]. Retrieved from https://mostlymktg.com/citation-index/[slug] Licensed CC BY 4.0 — https://creativecommons.org/licenses/by/4.0/
Academic (APA-ish): Mostly Marketing. (2026). AEO Citation Index [Dataset]. https://mostlymktg.com/citation-index
References
Sources & further reading
- Generative AI and the future of search — Google Search Central Blog, 2024.
Google's own description of how AI Overviews assemble and cite sources.
- How Perplexity works — sources and citations — Perplexity AI, 2024.
Vendor documentation for how live-web retrieval and per-sentence citations are produced.
- OpenAI. GPT-4 Technical Report — arXiv:2303.08774, 2023.
Establishes the stochastic nature of LLM outputs — the reason we run each prompt 4–5×.
- Structured Data General Guidelines — Google Search Central, 2024.
Basis for the schema.org markup we recommend for entity clarity.
- Creative Commons Attribution 4.0 International (CC BY 4.0) — Creative Commons, 2013.
License under which the Citation Index dataset is published.