State of AI Search: the engines disagree.

When buyers ask AI for the best tool in a category, the answer depends on which AI they ask — and the sources behind it are mostly not the vendor's own site. A measured look at share of model across developer-tool categories, reproducibly, with confidence intervals.

Method Median of 10+ runs/prompt · 5 engines · Wilson 95% CI Window June 2026 Updated 2026-06-20

Key findings

What the data shows

~95%

of AI citations are third-party

Across AI engines, roughly 95% of the citations behind answers come from third-party pages, not a brand's own site (Otterly, State of AI Search). Being cited where AI looks matters more than another homepage tweak.

engines, and they disagree

No single vendor leads on every engine for the same buyer prompt. Strong visibility on one engine says nothing about another, so we measure and report each engine separately.

10+

runs per engine, per prompt

AI answers are non-deterministic — a single screenshot is one draw from a distribution. We report the median across many runs with a Wilson 95% confidence interval.

categories measured live so far

CI/CD platforms and AI observability tools are live measured editions; the rest of the Index ships as clearly-labelled illustrative previews until their first live run.

Benchmarks

Share of model by category

The % of qualifying answers that name each product — median across five engines, with a Wilson 95% confidence interval. Share-of-model only; not SEO rankings.

CI/CD platforms

10 prompts · 5 engines

#	Product	SoM	95% CI
1	GitHub Actions	21.3%	19.4–23.4%
2	GitLab CI/CD	19.7%	17.8–21.7%
3	CircleCI	18.6%	16.7–20.5%
4	Jenkins	16.5%	14.7–18.3%

AI observability tools

10 prompts · 5 engines

#	Product	SoM	95% CI
1	Datadog	17.7%	16.1–19.5%
2	New Relic	14.9%	13.4–16.6%
3	OpenTelemetry	14.8%	13.3–16.4%
4	Grafana	13.6%	12.2–15.2%

Measured snapshot (June 2026) from our AI Visibility Index — median of 10+ runs/engine across 5 engines, Wilson 95% CI. Four further categories (API platforms, databases, feature flags, vector databases) are illustrative previews until their first live run — see the Index for the full set.

Method & honesty

How we measured this

Reproducible, with caveats. Share of model is the % of qualifying answers naming a product, measured as the median of 10+ runs per buyer prompt across Perplexity, Google Gemini, ChatGPT (OpenAI), Claude (Anthropic) and Grok (xAI), with a Wilson 95% confidence interval. We publish the date and method with every figure, and we don't publish numbers we haven't measured. AI answers are non-deterministic and change over time; this is a point-in-time measurement, not a guarantee. The two leaderboards above are live measured editions; remaining categories are clearly labelled illustrative previews.

Cite this report

Reference

Suggested citation

Clear Cited. "State of AI Search: the engines disagree." 2026. clearcited.com/research/state-of-ai-search

Journalists: for the underlying figures or written commentary, email hello@clearcited.com.