AI Visibility Index: which sources AI answers actually cite

Baseline collected October 4, 2026 (UTC). 15 questions × 3 independent runs = 45 grounded answers. Machine-readable copy: download the dataset (JSON).

What this is

FoundNow measures whether local businesses show up in AI answers. When an assistant answers, it also cites the sources it relied on - and those citations decide which sites get quoted, recommended and linked across the AI answer layer. This page is a public, reproducible record of which domains actually get cited when an AI assistant answers the questions local-service owners ask about AI visibility (AEO/GEO).

We publish the full citation pool, the exact questions, the exact method, and our own score - including our own zero. Every rate on this page carries its sample size. No figure on this page is a promise about outcomes; it is a measurement of one snapshot in time.

Headline numbers

Read the second row literally: in 45 grounded runs, the assistant never cited our own domain. With a true rate of 0 observed in 45 runs, the Wilson 95% interval runs from 0 to 0.079 - the data is compatible with any underlying rate below roughly 7.9%, and with zero. We report the interval, not just the point estimate, because that is the standard we sell.

Method (reproducible)

• Engine: Gemini, model string gemini-3.5-flash-lite, via the Google AI Studio web interface, grounded with Google Search (free tier — total spend: $0).
• Questions: 15 fixed questions about AI visibility for local service businesses (plumbers, electricians, law firms, dentists, multi-location brands). Exact text below.
• Runs: 3 independent runs per question, each in a fresh chat. 45 runs total.
• Recorded per run: the full answer, the cited source domains, and whether our own domain appeared.
• Citation counting: each answer carries a de-duplicated list of cited domains. A domain's figure is the number of runs (out of 45) whose citation list contained it.
• Date: October 4, 2026 (UTC). All 45 answers returned numbered citations; none failed.

Limits, stated plainly: this is one engine surface (Gemini with Google Search grounding), one question set, one collection day. Three runs per question is enough to describe the citation pool, not to rank any single domain within one question. The aggregate pool (45 runs, 422 citation observations) is the stable part of this dataset; treat single-question readings as indicative. No personalization was applied - fresh chats, no history.

The 15 questions

Fixed question set, asked verbatim in every run:

1. Q1. Search the web: what are the top 3 AI visibility (AEO) tracking tools for local service businesses? List each with its source URL. (unique domains cited: 16)
2. Q2. Search the web: which platforms help small local businesses (plumbers, electricians, lawyers) show up in ChatGPT and AI Overviews? Name the tools and cite sources. (unique domains cited: 23)
3. Q3. Search the web: best AI search visibility software for multi-location service brands in 2026. List tools with source URLs. (unique domains cited: 20)
4. Q4. Search the web: I run a local home service business. How do I get my business recommended by AI assistants? Which paid tools actually work? Cite sources. (unique domains cited: 33)
5. Q5. Search the web: what does "Answer Engine Optimization" software cost, and which vendors offer it? Cite sources. (unique domains cited: 14)
6. Q6. Search the web: compare Profound, Peec AI, Scrunch AI, and other AEO tools for tracking brand mentions in AI answers. Cite sources. (unique domains cited: 19)
7. Q7. Search the web: which AEO tools track how often a local business is cited by ChatGPT, Gemini, and Perplexity? List them with sources. (unique domains cited: 10)
8. Q8. Search the web: best GEO (Generative Engine Optimization) tools for a local services marketplace in the United States. Cite sources. (unique domains cited: 19)
9. Q9. Search the web: what agencies or freelancers use to audit AI visibility for local businesses before running ads? Cite sources. (unique domains cited: 21)
10. Q10. Search the web: is there a tool that shows which websites mention my business when AI answers local service questions? Cite sources. (unique domains cited: 25)
11. Q11. Search the web: recommend an affordable AEO platform for a small agency serving local service clients. Name at least 3 options with source URLs. (unique domains cited: 17)
12. Q12. Search the web: how do I track AI Overviews visibility for a dental clinic or law firm? Which software is used? Cite sources. (unique domains cited: 24)
13. Q13. Search the web: top 10 AI search monitoring tools in 2026 with pricing, especially ones aimed at SMB service businesses. Cite sources. (unique domains cited: 21)
14. Q14. Search the web: does ChatGPT search or Perplexity have a directory of businesses like Yelp? Which AI tools index local providers? Cite sources. (unique domains cited: 18)
15. Q15. Search the web: what software identifies which third-party websites get cited by AI for local service queries? Cite sources. (unique domains cited: 20)

Per-question results

"Unique domains" counts distinct domains cited across that question's 3 runs. "Most-cited" lists the domains cited in the most of those runs (×n = cited in n of the 3 runs).

Most-cited domains (top 20)

Ranked by the number of the 45 runs in which the domain appeared in the answer's citation list.

The shape of the top of this table is the actionable part: the most-cited source overall is youtube.com (16/45 runs), followed by the vendor site tryprofound.com (13/45), reddit.com (12/45) and dageno.ai (12/45). Search-industry references (semrush.com 11/45, searchenginejournal.com 4/45) and a regional news site (inquirer.net 11/45) sit in the same band as the AEO vendors themselves. Individual practitioner sites appear too - nicklafferty.com was cited in 10/45 runs, more than most venture-backed tools in this list.

Full citation leaderboard (187 domains)

Every domain observed in the citation lists of the 45 runs, sorted by the number of runs that cited it. The same data is in the JSON dataset.

AEO/GEO vendor domains in the citation pool

These are domains of vendors selling AI-visibility or generative-engine optimization tracking - our category. The classification is ours; the counts come from the same 45 runs as every other figure on this page.

What the citation pool looks like

• Video and community sources lead: youtube.com (16/45) and reddit.com (12/45) are the two most-cited non-vendor domains in the pool.
• Practitioner and niche blogs are heavily represented: 187 domains were cited in total; 84 of them appeared in more than one run, and a single practitioner blog (nicklafferty.com, 10/45) outranks most funded vendors.
• Regional press is in the trust pool: inquirer.net was cited in 11/45 runs - as often as semrush.com (11/45).
• Review and directory sites appear steadily: technologyadvice.com (6/45), zapier.com (6/45), hubspot.com (6/45), g2.com (3/45).
• The long tail is long: 103 domains were cited in exactly 1 of 45 runs. Being cited once is common; being cited repeatedly is what separates the top of this table.

Why we publish our own zero

We sell AI-visibility measurement. At this baseline, our own domain was cited in 0/45 runs (Wilson 95% CI [0, 0.079]). We publish that number with its interval because our product's rule - no rate without a sample size, no single-value claims - applies to us first. The gap this page documents is the gap our customers use the product to close: you cannot manage whether AI answers cite you until you can measure it.

Reproduce and update

• The question set, engine, model string and counting rule are fixed (above).
• The full per-run dataset behind every figure on this page is the machine-readable copy: ai-visibility-index-2026-10.json (per-question citation lists included).
• We re-run this benchmark on a weekly cadence with the same method and publish the next dated snapshot; comparisons between snapshots use the same interval math as everything else we publish.
• If you reference this data, link to this page and include the snapshot date (October 4, 2026).

Related

Full sampling methodology (product tiers, Wilson intervals, model-version tracking): Methodology. Short version: FAQ. Want this measurement for your own business? Run the one-time scan or see pricing - and note what we do not do: no placement promises, no single-run findings. Questions: contact us.