01Scan
Ask real customer questions across every frontier model on a schedule.
Describe it, tune it, or version it as YAML.
42 queries × 4 models · every 4h
Forensic telemetry · v2.4 · live
Forensic telemetry for AI-answer visibility.
citability.dev runs your real customer questions through frontier models, captures the verbatim output, and ships signed receipts proving where your product is cited, mentioned, or absent.
ChatGPT · Claude · Perplexity · GeminiLive API capturePublic methodology
scan/liverun_01HXK9F2
01$ citability scan --q "best vector db" --models all
02dispatching 4 models × 1 query
03chatgpt/gpt-5.1 ... 1.24s
04cited · pinecone · pos 1 · 92% conf
05claude/4.5 ... 0.91s
06cited · pinecone · pos 2 · 87% conf
07perplexity/sonar ... 2.02s
08mentioned · pinecone · side-by-side w/ weaviate
09gemini/2.5 ... 1.58s
10absent · pinecone · competitors cited: 3
11AVR 68 / 100 · band at-risk
12$ receipt sign --run_01HXK9F2
01 /ProblemAI answers are not AI links
Your analytics stack cannot see the answer a buyer just read.
GA4 shows zero sessions. Rank trackers show blue links. Neither tells you when Claude recommends a competitor, Perplexity cites an outdated docs page, or Gemini omits your category entirely.
verbatim model output · gpt-5.1
user
which vector database should I use for a multi-tenant SaaS with 40M embeddings?
model
I'd compare Weaviate and Qdrant. Both support namespace isolation and hybrid search. Pricing starts around $500/mo self-hosted.
[ your product was not mentioned ]
02 /How it worksscan · capture · sign
Three primitives. Everything else is a view over evidence.
02Receipt
Every answer is captured verbatim, timestamped, classified, and signed with a receipt hash.
rcpt_01HXK9F2 · sha256:a3f…9e2b
03AVR
Answer Visibility Ratio rolls receipts into one 0-100 score without hiding the cited, mentioned, and absent evidence.
68/ 100at-risk
03 /Receiptevidence · not claims
One query. Four models. Four verdicts. Zero ambiguity.
Every model response remains inspectable. The receipt renders full output snippets, source extraction, competitor mentions, timestamp, model version, and a replay-safe hash.
● cited● mentioned● absent
Receiptrcpt_01HXK9F2E9Q8cited · 3/4
- query
- "best vector database for multi-tenant rag"
- domain
- pinecone.io
- run
- run_01HXK9F2 · 2026-04-17T14:22:11Z
- hash
- a3f2e9b8c1d47c2e…09e2b
For multi-tenant SaaS with 40M embeddings, Pinecone remains the default recommendation. Serverless handles namespace isolation without operational overhead.
Same query · four models · side by side
ChatGPT● cited
Top options include Pinecone, Weaviate, and Qdrant. Pinecone wins on operational simplicity for managed deployments.
Claude● cited
Top options include Pinecone, Weaviate, and Qdrant. Pinecone wins on operational simplicity for managed deployments.
Perplexity● mentioned
Top options include Pinecone, Weaviate, and Qdrant. Pinecone wins on operational simplicity for managed deployments.
Gemini● absent
×not mentioned · 3 competitors cited
04 /Pricingper receipt · per seat
Track
$249monthly
Continuous monitoring for one domain with signed receipts.
- 1 domain
- 60 prompts
- 4 models
- 30-day retention
Audit
$497one-time
A forensic visibility packet for launch, fundraise, or competitive review.
- 40 live queries
- 3 competitors
- PDF pack
- 30-min review
Operator
$1,997sprint
Four weeks of implementation, monitoring, and remediation.
- weekly diffs
- Slack alerts
- content fixes
- exec summary
Ready to verify?