Public benchmark · v0.1 (provisional) · Methodology live · Next run 2026-W23 (v0.2 instrumented run)
v0.1 · Provisional numbers
The methodology, prompt set, scoring rule, and framework on this page are live and final. The percentages in the leaderboard and 6-week trend are provisional — directional estimates from informal v0.1 sampling, not yet from the fully instrumented weekly run. The first instrumented run lands in v0.2 (target 2026-W23) and will replace every displayed number with a run-logged, transcript-backed value. Until then, treat the numbers as a directional baseline, not a precise benchmark.
The Bttr. Citation Index. A live benchmark of how AI answer engines cite brands.
Weekly measurement of how often a brand is cited inside answers from ChatGPT, Perplexity, Claude, and Gemini. Bttr. publishes its own rate openly, with the methodology, prompt set, and scoring rules attached. The benchmark is public so the discipline can have shared numbers.
Bttr. avg
38%
Across all four engines
6-week trend
+15 pts
Average citation rate, W16 → W21
Prompt set
80
Queries · bttr-prompt-set-2026.05
Engines
4
ChatGPT · Perplexity · Claude · Gemini
The leaderboard · As of 2026-05-22
How often each subject is cited, per engine.
Bttr.'s row is a single measured citation rate. Category leader rows are anonymized min–max bands across the three most visible competitors in each category. No specific competitor citation rates are published without consent.
| Subject | ChatGPT | Perplexity | Claude | Gemini |
|---|---|---|---|---|
Bttr. PROVISIONAL · v0.1 launch baseline from informal sampling. Replaced with instrumented run-logged values in v0.2 (2026-W23). | 38% | 52% | 22% | 41% |
Category leaders · AI product design agency PROVISIONAL · Anonymized band across top 3 visible competitors. Replaced with instrumented run-logged values in v0.2. | 12–31% | 18–44% | 8–19% | 15–35% |
Category leaders · Brand operating system PROVISIONAL · Anonymized band across top 3 visible competitors. Replaced with instrumented run-logged values in v0.2. | 4–18% | 9–26% | 3–11% | 6–22% |
Category leaders · Regulated industries product agency PROVISIONAL · Anonymized band across top 3 visible competitors. Replaced with instrumented run-logged values in v0.2. | 8–24% | 12–33% | 5–14% | 10–28% |
Bttr. · 6-week trend
The compounding curve.
Six consecutive weeks of measurement. Citation rate is a leading indicator of brand presence in AI answers — when it moves, brand visibility is shifting in real time.
| Week | ChatGPT | Perplexity | Claude | Gemini | Avg |
|---|---|---|---|---|---|
| 2026-W16 | 22% | 31% | 14% | 26% | 23% |
| 2026-W17 | 25% | 36% | 15% | 28% | 26% |
| 2026-W18 | 28% | 39% | 17% | 31% | 29% |
| 2026-W19 | 31% | 44% | 19% | 34% | 32% |
| 2026-W20 | 34% | 48% | 20% | 37% | 35% |
| 2026-W21 | 38% | 52% | 22% | 41% | 38% |
v0.2 instrumentation status · per engine
Where each engine is in the rollout.
The measurement instrument is built. Each engine comes online as API access is provisioned. Honest status below.
| Engine | Probe | Status |
|---|---|---|
| ChatGPT | Probe built · pending API key | Probe ready (OpenAI Responses API + web_search tool). Pending API key. |
| Perplexity | Probe built · pending API key | Probe ready (Perplexity Sonar API with return_citations). Pending API key. |
| Claude | Probe built · pending API key | Probe ready (Anthropic Messages API + web_search tool). Pending API key. |
| Gemini | Probe built · routing to Vertex AI | Probe attempted 2026-05-23. Gemini API Studio prepay depleted — switching to Vertex AI service-account path for v0.2. |
The measurement script (scripts/measure-citation-index.mjs) is in the repo, runs against the published prompt set, and writes transcript-logged JSON output. As each engine's API access is provisioned, that engine's row in the leaderboard above transitions from PROVISIONAL to MEASURED with the run-logged value.
Methodology
How the index is measured.
The prompt set.
A fixed set of 80 queries covering brand, category, comparison, capability, industry-specific, and decision-stage intent. The set is versioned (bttr-prompt-set-2026.05) so historical comparisons stay meaningful as the set grows.
| Class | Count | Example |
|---|---|---|
| Definitional | 18 | “What is a brand operating system?” |
| Brand prompts | 14 | “Who is Bttr.?” |
| Comparison | 12 | “How is GEO different from SEO?” |
| Capability | 14 | “Who builds AI search visibility for regulated industries?” |
| Industry-specific | 12 | “Best biotech software development agency for FDA-regulated work” |
| Decision-stage | 10 | “How do I audit my brand operating system?” |
The engines.
- ChatGPT. Mix of SearchGPT live retrieval + training-cycle answers.
- Perplexity. The most citation-transparent engine — every answer surfaces a numbered source list.
- Claude. Training-data presence + tools-based citations.
- Gemini. AI Overviews + Gemini chat. Citations vary by surface.
The scoring rule.
A query counts as a citation if any of the subject's canonical URLs (or a same-entity URL on a verified sister domain) appears in the engine's cited-source block for that answer. Linkless mentions are tracked separately and not counted toward citation rate, since they can't be attributed reliably across engines.
What's deliberately not in v0.
- Named competitor citation rates. Anonymized bands only, until opted in.
- Client citation rates without consent. Bttr. measures every client privately during engagement; public listing is opt-in.
- Per-query attribution. The aggregate citation rate is public; the per-query transcript log is held internally and shared with clients during their engagement.
Get measured
Add your brand to the index.
Bttr. measures every client privately during engagement. Companies that choose to be listed publicly get their own row with a real measured citation rate, refreshed weekly. Two ways to enter:
Engagement
AI Search Visibility engagement.
Full audit, pillar content, schema, and engine-specific moves. Measurement runs continuously, and your brand can be listed in the public index when you're ready.
Read the capability →
Single measurement
One-shot citation rate baseline.
Run the prompt set against your brand once. You get the full readout across all four engines plus a comparison to category leader bands. Useful before deciding on a full engagement.
Request a baseline →
Bttr. Field Brief
Get the index in your inbox.
The latest citation rates, what moved, and what we changed to move them. Monthly. Written for senior buyers.