What counts as an AI citation in this study?

An AI citation is an AI answer that links to a client page as a source. Mentions, where the brand is named without a link, are tracked separately and never added to this count. Each citation is counted once after de-duplication across the three tools.

Does ranking on page one cause the AI citation?

Ranking does not cause the citation, but retrieval comes first: an engine can only quote a page it can fetch and trust. In the case files the technical and Internal Linking work preceded the citation growth; the order of events is shown, not a claim of cause.

Why is there no comparison with other agencies or peer sites?

The peer comparison was cut because two of 24 peer domains caused 89.5% of the peer set's decline while the median peer grew 41.8%. A headline built on that total would have misread the typical site, so the study reports the client portfolio on its own.

Why does one client hold half the total?

One client, a printing company with five country storefronts, holds 48.6% of the citations because each storefront is counted as its own site in its own market. Growth without that client is 88%, and the median client grew 45%, so the headline does not depend on it.

Can SERP Masters run this measurement on my site?

Yes. The AI Visibility Audit runs your site through the same three tools, reports each engine on its own line, and lists where competitors are cited instead of you. It is the first deliverable of every AI SEO engagement.

AEO Case Study · 28 Sites · 3 Measurement Tools

AI Citation Case Study

Q: Why use three tools instead of one?

Three tools were used because no single tool sees every citation. In this study only 18.9% of one tool's AI Overview keywords appeared on the other's list, and 87 of 307 live renders found citations the first tool had missed. One tool gives a floor, not a count.

15,332 AI citations across 28 client sites. Up 80% in a year, on the same engines, pulled on the same day.

By Andy Wang · founder of SERP MastersB.S. Pharmacology, UC Santa Barbara · Pokémon, Disney, Mastercard, Mailchimp · About Andy →

This AI citation case study counts every AI citation across 28 SERP Masters client sites with three independent tools, removes the duplicates, and compares the same engines a year apart. SERP Masters is an SEO agency run by founder Andy Wang. The portfolio holds 15,332 de-duplicated AI citations across Google AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini and Copilot. Like-for-like growth is 80% year over year, from 7,642 to 13,755. Five of the six sites that had zero AI citations a year earlier are cited today, and 87 of 307 live AI Overviews cited the client where one tool showed no citation.

Request an AI Visibility Audit

Summary5 points

15,332 AI citations detected across 28 client sites, each counted once after de-duplication across Ahrefs, DataForSEO and live renders.

Like-for-like growth of 80% year over year on the same engines, 7,642 to 13,755. Without the largest client, 88%. Median client, 45%.

5 of 6 sites with zero AI citations a year ago are cited today.

87 of 307 high-volume AI Overviews marked “not cited” by one tool cited the client when rendered live.

No single tool sees every citation: only 18.9% of one tool’s AI Overview keywords appeared on the other’s list.

Most AEO case studies show one screenshot from one tool with no baseline. This one shows every number’s source, the formula behind every derived figure, and the figures SERP Masters chose not to publish.

The study is the measurement layer of the AI SEO services engagement: the same prompt panel and the same three tools run on every client account, every month.

What the study counted

The study counted AI citations: an AI answer that links to a client page as a source. A citation is not a mention. A mention names the brand without a link, and SERP Masters tracks mentions on a separate line, never averaged into this count.

FigureValueHow it was counted

AI citations across the portfolio15,332Ahrefs 14,755 (all tracked engines) plus 4,856 DataForSEO-only AI Overview keywords plus 60 live-render-only citations, minus 4,339 for possible overlap

Year-over-year growth, same engines+80%7,642 to 13,755 across AI Overview keywords, ChatGPT, Perplexity, Gemini and Copilot, both years pulled on the same day

Growth without the largest client+88%The largest client, a printing company with five country storefronts, holds 48.6% of the total

Median client growth+45%The middle client in the year-over-year distribution

AI Overview keywords only+87%6,664 to 12,479

Zero-to-cited sites5 of 6Sites with zero AI citations a year ago that are cited today

Live AI Overview check87 of 307High-volume AI Overviews with no citation in one tool that cited the client when rendered live (28.3% of 307; 324 keywords checked)

Tool overlap18.9%1,128 of 5,984 DataForSEO AI Overview keywords also appeared on the Ahrefs list

Every figure above is Measured (straight from a tool) or Derived (computed by the stated formula). None is projected.

Finding one: no single tool sees every citation

No single AI visibility tool sees every citation. The tools sample different keyword and prompt sets, and AI answers pull from pages far beyond the organic top 10, so two tools run on the same 28 sites overlapped on fewer than one in five AI Overview keywords.

01Tool one: Ahrefs.Every engine it tracks: AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini and Copilot. 14,755 citations across the 28 sites.

02Tool two: DataForSEO.Google AI Overview keywords citing the client that the first tool did not have. 4,856 keywords added only where the exact keyword was missing from tool one.

03Tool three: live renders.324 high-volume keywords rendered in a real browser, geotargeted for local clients. 60 citations appeared in neither database, and 87 of 307 AI Overviews cited the client where tool one showed none.

04Minus overlap: 4,339.Wherever the first tool reported more AI Overview keywords than it listed, the study assumed the second tool’s finds were already among them, so the total errs low.

The practical consequence for a client report: a single-tool AI visibility score understates citations and, worse, understates them unevenly by engine. SERP Masters reports each engine on its own line and names the tool beside every count.

Finding two: the sites that went from zero to cited

Five of the six sites with zero AI citations a year ago are cited today. The sixth, a church site, still has none. The five that moved share one thing: the technical work that let search and AI crawlers reach the pages came first, and the citations followed it.

Site, anonymizedAI citations a year agoAI citations today

Behavioral health network, multi-facility0719

B2B software company043

Behavioral health provider028

Specialty ecommerce store013

Transportation service02

Church site00

The order of events is shown, not a claim of cause. AI citations rotate often, and a count from one tool on one day is a detection, not a position.

Case files: what shipped, and what the citations did next

Months are counted from the start of each engagement. Citation counts compare the same engines a year ago and today.

Evidence · DTC supplement brand, Shopify1616 to 344 AI citations in one tool (393 across all three); Gemini alone from 1 to 111.

Client or vertical · Direct-to-consumer supplement brand on Shopify.Baseline · 16 AI citations a year ago; Gemini cited the site once.Intervention · Month 1, technical audit. Month 2, 200+ redundant hreflang tags removed, collection meta rewritten, Internal Linking across 56 articles in 8 topic silos. No AI-specific work.

Read the case study →Evidence · behavioral health network, multi-facility00 to 719 AI citations in one tool (721 across all three).

Client or vertical · Multi-facility behavioral health network, YMYL healthcare.Baseline · Zero AI citations in any month before the technical work went live.Intervention · Month 1, technical audit and redirect map. Month 3, server-level fixes, robots.txt corrected, on-page optimization.

Read the case study →Evidence · industrial test equipment, ecommerce99 to 220 AI citations in one tool (345 across all three).

Client or vertical · Branded catalog of industrial test instruments.Baseline · 9 AI citations a year ago.Intervention · Month 1, technical audit, fix list and JSON-LD schema implementation. Month 2, 122 category pages rewritten with Internal Linking across a three-tier category tree.

Read the case study →

What the three case files have in common: none of the work was labeled “AEO”. Crawl access, redirects, schema, Internal Linking and answer-first rewrites of pages that already ranked moved the citation counts, which is why SERP Masters sells AI search work on top of the technical and content foundation rather than instead of it.

The figures SERP Masters chose not to publish

Three figures came out of the study and were left out of every public page, because the verifier could not stand behind them.

01A 96.2% share for one tool.It is an output of the de-duplication rule, not a measurement, so it says nothing about the tool.

02A peer comparison.Twenty-four non-client domains were tracked beside the portfolio. Two of them caused 89.5% of the peer set’s year-over-year drop, the median peer rose 41.8% and 16 of 24 peers grew. A headline that read “clients up, peers down” would have been true of the total and false of the typical peer, so the comparison was cut.

03“Up from 15,332.”An earlier run of the same study produced 15,332. The tool has since restated its history for two of the largest sites, so the two totals are not comparable and the earlier figure is retired rather than used as a baseline.

Publishing the method and the rejected figures is the point of the study. An AI visibility claim without a stated denominator, tool and window is a screenshot, not a measurement.

How the study was run

The study was run the same way SERP Masters measures every client account, so a client’s monthly report and this page use one method.

01A year apart, same day, every engine.AI citation counts for all 28 sites across AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini and Copilot, with both years pulled on the same day so tool changes cannot skew the comparison.

02A second tool, matched keyword by keyword.DataForSEO AI Overview references in each site’s main market, added only when the exact keyword was missing from the first tool.

03Live renders, city by city.324 high-volume keywords rendered in a real browser, geotargeted for local clients, to catch citations neither database had. The pool is a sample of top-volume keywords at best position 10 or better, filled to 30 or better where needed; it is not “page one”, and the 28.3% rate is not applied to the total.

04Conservative de-duplication.Where two tools might have counted the same citation, the study assumes they did, so the total errs low.

05Measured or derived, never guessed.Each figure is labeled Measured (straight from a tool) or Derived (computed by a stated formula). No projections.

The work behind the citations

The work behind the citations is the same foundation that moves rankings, with the answer layer on top.

Technical SEO and AI crawler access.Clean crawl paths, correct redirects and indexation, and AI retrieval bots allowed through. See technical SEO services.

Internal Linking and topical structure.Topic silos that tell search engines and AI systems which page answers which question.

JSON-LD schema implementation.Organization, local business and procedure markup that ties the brand to its entities.

E-E-A-T and entity work.Credentials, author and About pages, and consistent third-party profiles for YMYL brands.

Answer-first content.Pages that answer the question in the first lines, the format AI answers quote. See answer engine optimization services.

Multi-tool AI visibility tracking.Monthly counts across tools and engines, de-duplicated and tracked over time, reported per engine.