Is the daily scrape automated or manual?

Automated. A query harness runs each engine daily through documented public endpoints and (where direct API access is not yet available) through robust headless-browser captures. Every capture is timestamped and the raw response stored.

Which AI engines are tracked?

Engine coverage is selected at kickoff from ChatGPT, Perplexity, Gemini, Claude, Grok, Google AI Mode, and Google AI Overviews. Engagements typically lock 4 engines for the 45-day measurement window. The set cannot change mid-window.

How are queries selected?

Together. KailxLabs proposes 20 high-intent buyer prompts based on the vertical and city; the client reviews and ratifies. Queries must be ones a real prospect would type unaided. Queries that contain the client business name are excluded — only category-level intent counts.

What is the day-45 citation report?

A document delivered to the client at day 45 containing the 20-query set, the 4-engine set, daily capture timestamps, citation counts per query per engine, the raw response text for every qualifying citation, and a refund determination (cited or not cited by the public definition).

Can a client see citation status before day 45?

Yes. Most engagements receive a Day-14 interim snapshot and a Day-30 progress check. The day-45 report is the formal measurement, but early signal is shared.

What if an AI engine introduces personalization that affects scrape consistency?

Captures run from clean, geo-targeted sessions matched to the client's city. Logged-out, no chat history, no fingerprint persistence. This is documented in the day-45 report so the client can verify the scrape conditions match real cold-buyer behavior.

Are the raw logs publishable as a case study?

Only with explicit client consent. Default behavior is internal use plus client audit. Redacted versions can be published as anonymized proof when the client approves.

How we measure AI visibility — the KailxLabs daily scrape methodology

KailxLabs sells an AI Citation Foundation Build with a 45-day citation guarantee. The build is the product; the daily measurement harness is the proof. This page documents how AI visibility is measured during the 45-day window so the guarantee is verifiable, not narrative.

The harness in one paragraph

Short answer. An automated daily harness runs the 20 agreed buyer-intent prompts across the 4 agreed AI engines from clean geo-targeted sessions. Each response is captured in full, timestamped, and stored as raw text. The day-45 citation report contains every capture and a determination of whether the engagement met the guarantee criteria defined on /what-counts-as-a-citation.

What the harness captures

Short answer. Three things: the full text response from each AI engine for each query, the timestamp of capture, and the session conditions (engine version where available, geo, user-agent, logged-out state).

The raw capture is the source of truth. Every claim in the day-45 report links back to a specific capture timestamp. The client can read the underlying AI response for any cited query — no aggregated chart that hides the source data.

Engine coverage

Short answer. ChatGPT, Perplexity, Gemini, Claude, Grok, Google AI Mode, and Google AI Overviews are all supported. Each engagement selects 4 engines at kickoff. The locked set is what the day-45 guarantee evaluates.

Most clinic and law engagements lock ChatGPT + Perplexity + Gemini + Google AI Overviews. Med-spa and home-services engagements often substitute Claude in place of one of the four when the vertical\'s buyer behavior favors a specific engine. The choice is documented before the build starts.

Query selection

Short answer. KailxLabs proposes 20 high-intent buyer prompts based on vertical, city, and procedure or service mix. The client ratifies. Queries that contain the client business name are excluded — only category-level intent qualifies. The 20 queries are locked at kickoff.

Three rules constrain the query set:

Cold prompts only. The buyer journey we measure is the prompt a real prospect would type. No clarifiers, no narrowing.
No branded queries. "Is KailxLabs reputable" is not a category citation. The set tests category discovery, not brand verification.
Vertical-realistic. The queries reflect how patients, clients, or homeowners actually phrase the search. Not how SEO consultants phrase it. Plain English.

Geo and session conditions

Short answer. Sessions run from clean, logged-out, geo-targeted contexts matched to the client\'s primary city. No chat history. No user fingerprint persistence. This isolates the AI\'s organic retrieval behavior from personalization noise.

The client\'s primary city is locked at kickoff. If a clinic operates in Austin, all 20 queries run as if asked from an Austin user. AI engines that detect geo from IP plus query context will respond as they would to a real Austin buyer.

Interim signal

Short answer. A Day-14 snapshot and a Day-30 progress check are shared with the client during the 45-day window. The day-45 report is the formal measurement, but the client sees signal as it accumulates.

Day-14 is typically when the first citations land. Day-30 is the compounding curve. Day-45 is the binary determination for the guarantee. Most engagements that will hit the guarantee have hit it by day 30; the remaining 15 days are buffer plus stability check.

The day-45 citation report

Short answer. A single document delivered to the client containing the 20-query set, the 4-engine set, daily capture timestamps for all 80 query-engine slots × 45 days, citation counts per query per engine, raw response text for every qualifying citation, and a clear refund determination (cited or not cited per the public definition).

The report is the artifact the guarantee runs on. If the determination is "not cited," the refund processes within 7 days. The client keeps the website, code, schema, and content regardless.

Publishability

Default behavior: the day-45 report is for internal use and client audit. The client may approve a redacted version for use as anonymized proof (name removed, city kept, query examples kept, citation counts kept). Some current case studies on kailxlabs.co/case-studies are this kind of redacted proof. Named-client proof requires explicit written consent and is rarer.

What this page does not promise

It does not promise the harness captures every transient AI behavior. AI engines change retrieval logic; the harness captures behavior at the time of each scrape, not a Platonic ground truth. It does not promise the daily scrape catches every micro-spike. A citation that appears between scrape runs and disappears is not captured. The daily cadence is calibrated to capture stable citations; ephemeral ones are not the guarantee target.

What counts as a citation — the public definition the daily scrape evaluates against. The full KailxLabs methodology — how citation-worthy pages are built before measurement begins. 40-Clinic AI Visibility Audit (Q1 2026) — the same harness applied to a research dataset, published under CC BY 4.0.