AI search engine overlap and query volatility

What local-result overlap, domain diversity, language shifts, and paraphrases reveal about the instability of AI search visibility.

By dotSuper Research DeskPublished Sep 1, 2026Reviewed Sep 1, 202610 min read
Search & discoveryOverlap and robustness experiments from arXiv:2509.08919Updated Sep 1, 2026

/ THE SHORT ANSWER

It can be highly unstable. In the sampled local-service queries, overlap between Google’s top results and AI-cited domains ranged from 20.6% for home cleaning to 0.1% for IT support. Cross-engine domain overlap was also low, while language changes had a larger effect than simple paraphrasing.

Key takeaways
  • 01The Google-to-AI overlap in local-service queries fell below 3% for auto repair and IT support in the paper’s sample.
  • 02Automotive citation sets contained 350 distinct Claude domains, 347 Perplexity domains, and 212 ChatGPT domains, with low pairwise Jaccard overlap.
  • 03Language shifts changed source domains more than prompt paraphrases in the tested query set.
  • 04A useful dashboard must retain engine, locale, prompt, and capture date instead of compressing everything into one GEO score.

/ dotSuper point of view

Visibility is a distribution, not a rank. Measure the query set, engine, market, language, and date together.

Google and AI search can surface very different local domains

The paper compared Google top results with AI-cited domains for six local-service categories. Even the highest overlap, home cleaning at 20.6%, leaves most domains unmatched between the two surfaces.

LOCAL DISCOVERY

Overlap falls from one in five to almost zero.

Share of domains appearing in both Google results and AI citations for the sampled local-service queries.
Home cleaning20.6%
Roofing17.1%
Tax preparation15.4%
Dentists11.9%
Auto repair2.5%
IT support0.1%

The paper reports category-level overlap for its sampled prompts. It does not establish the expected overlap for every city or service market.

View the chart data
Google and AI domain overlap for local services
Service categoryOverlap
Home cleaning20.6%
Roofing17.1%
Tax preparation15.4%
Dentists11.9%
Auto repair2.5%
IT support0.1%

Each engine builds a different citation neighborhood

For automotive prompts, Claude returned 350 distinct domains, Perplexity returned 347, and ChatGPT returned 212. Pairwise Jaccard overlap ranged from 0.096 to 0.251. Consumer electronics showed a similar pattern, with pairwise overlap from 0.088 to 0.200.

The practical consequence is simple. A page appearing in one engine does not establish broad AI visibility, and an engine-level win can disappear when the prompt, locale, or answer mode changes.

AUTOMOTIVE DOMAIN SETS

Similar questions, different source universes.

Distinct domains cited in the paper’s automotive experiment.
Claude350 domains
50.3% unique share
Perplexity347 domains
56.5% unique share
ChatGPT212 domains
60.8% unique share

A high unique share means many domains appeared only in that engine’s citation set.

View the chart data
Automotive citation-domain diversity
EngineDistinct domainsUnique share
Claude35050.3%
Perplexity34756.5%
ChatGPT21260.8%

Language matters more than small wording changes

Across the paper’s multilingual experiment, Google’s cross-language domain overlap was mostly between zero and 0.1. GPT showed near-zero cross-language overlap, while Claude reused more English-language authority sources. Gemini and Perplexity sat between those patterns.

Paraphrases were more stable. AI engine domain overlap often fell between 0.3 and 0.7, while Google was usually lower except for closely related imperative and keyword variants. This supports a market-language strategy, not a program of publishing one page for every wording variation.

How to model a visibility test
DimensionMinimum captureWhy it matters
EngineChatGPT, Claude, Gemini, Perplexity, Google AI featuresSource systems behave differently
IntentDecision, comparison, implementation, troubleshootingRetrieval changes with the job behind the query
LocaleCountry and language kept separateCross-language overlap can be very low
Prompt formNatural question plus two justified variantsTests robustness without manufacturing thin pages
TimeMonthly capture with model and dateResults are temporally unstable

What this page cannot conclude

  • 01Both papers are preprints. Their findings should be treated as evidence to test, not as a settled ranking formula.
  • 02The GEO engine study collected data in August 2025. Models, retrieval systems, interfaces, and citation behavior can change quickly.
  • 03Observed citation patterns do not prove that changing one page element will cause an engine to cite that page.
  • 04Google states that there is no special schema, file, or content format required for its generative AI features. Core SEO and useful, original content remain the foundation.

Sources

  1. 01Generative Engine Optimization: How to Dominate AI SearcharXiv · accessed Sep 1, 2026
  2. 02Optimizing your website for generative AI features on Google SearchGoogle Search Central · accessed Sep 1, 2026
BUILD A MEASURABLE DISCOVERY SYSTEMAI search engine overlap and query volatility

/ APPLY THE THINKING

Turn evidence into discoverable demand.

dotSuper connects original research, search foundations, AI visibility monitoring, content operations, and qualified lead routing as one accountable system.

Question for the working sessionHow stable is a brand’s AI search visibility across engines and query variants?

/ Topic-led working session · AI search engine overlap and query volatility

Turn this question\ninto a useful first move.

Bring how this question currently shows up in your business: “How stable is a brand’s AI search visibility across engines and query variants?” We’ll test the page’s evidence against your context and define the smallest useful next move.

Live availability from ceo@dotsuper.net Your time zone · Local time
  1. 01Bring the contextWhere this issue shows up in the work.
  2. 02Test the relevanceUse the evidence against your reality.
  3. 03Choose the next moveOne accountable action, clearly owned.
Live availability
  1. Date
  2. Time
  3. Booked

Syncing live times