Research honesty

Three tiers of claims — on every deliverable.

We label what is proven on file, what is directional, and what we refuse to say. COAST totals are diagnostic summaries of a defined test set — not probabilities of recommendation.

COAST Framework Limitations

Transparent measurement, inspectable evidence, and comparable reruns.
Measure → Fix → Verify
Full methodology and limits on dedicated pages — not repeated on every section.

How engagements work →

Research honesty

Three tiers of claims

We label what is proven on file, what is directional, and what we refuse to say — on every deliverable and sales conversation.

On the record

· Documented findings with inspectable evidence
· Verbatim excerpts with model + date
· Comparable verify rerun on the same questions

Directional

· Low-n exploratory sample — not a national survey
· Named list may shift between runs
· Accuracy and traceability use heuristic automation in v1

We do not claim

· Guaranteed AI rankings or map pack placement
· Survey-grade statistical certainty from one run
· Probability of recommendation from a single COAST total

Sources

Generative engine optimization benchmark literature — Conservative citation of published GEO/AEO research — benchmark lift ≠ client guarantee.
Google AI Overviews and retrieval documentation — Public documentation on how AI-generated summaries surface entities — not proprietary ranking logic.
OpenAI / Google public model behavior documentation — Model outputs vary by version, prompt, and retrieval context — we document all three on every run.

Solutions map each engagement step to artifacts with these honesty rules baked in. How we work →

Research honesty

Three tiers of claims — on every deliverable.

We label what is proven on file, what is directional, and what we refuse to say. COAST totals are diagnostic summaries of a defined test set — not probabilities of recommendation.

COAST Framework Limitations

Transparent measurement, inspectable evidence, and comparable reruns.
Measure → Fix → Verify
Full methodology and limits on dedicated pages — not repeated on every section.

How engagements work →

Research honesty

Three tiers of claims

We label what is proven on file, what is directional, and what we refuse to say — on every deliverable and sales conversation.

On the record

· Documented findings with inspectable evidence
· Verbatim excerpts with model + date
· Comparable verify rerun on the same questions

Directional

· Low-n exploratory sample — not a national survey
· Named list may shift between runs
· Accuracy and traceability use heuristic automation in v1

We do not claim

· Guaranteed AI rankings or map pack placement
· Survey-grade statistical certainty from one run
· Probability of recommendation from a single COAST total

Sources

Generative engine optimization benchmark literature — Conservative citation of published GEO/AEO research — benchmark lift ≠ client guarantee.
Google AI Overviews and retrieval documentation — Public documentation on how AI-generated summaries surface entities — not proprietary ranking logic.
OpenAI / Google public model behavior documentation — Model outputs vary by version, prompt, and retrieval context — we document all three on every run.

Solutions map each engagement step to artifacts with these honesty rules baked in. How we work →