What BoostGeo checks—and what each result means.

BoostGeo uses several kinds of evidence because no single score can prove whether an AI system will find, understand, cite, recommend, or successfully use a website. This page separates those methods so you can interpret every result at the right level.

Provider capabilities and availability can change. Each saved result records when the check ran, and supported product views identify the evidence and measurement mode used for that run.

Website Readiness

Question answered: Did BoostGeo retrieve representative public-page evidence, and how clearly does that page communicate important information?

Website Readiness applies deterministic checks to the public page BoostGeo retrieved. Depending on the page, these may include identity and offer clarity, headings and metadata, structured data, crawler access, trust information, and other page-type-relevant signals. The rubric is adjusted to the evidence and page type; a product page is not judged as though it were a local-service homepage.

• Evidence quality: Results identify whether representative evidence was available. Protected, unreachable, or materially unusable pages receive no numeric score.
• Rendered evidence: When an ordinary fetch is blocked, empty, or materially sparse, BoostGeo may use a bounded browser-rendering attempt. Persistent challenges, CAPTCHAs, login walls, and sensitive destinations are not bypassed.
• Meaning: The score describes the readiness of the assessed page evidence. It is not a security rating, whole-site grade, search ranking, or independent public AI-visibility measurement.

Optional model-comprehension checks

Question answered: Given the page information BoostGeo collected, how consistently do supported models interpret it?

These checks provide selected models with structured information collected from the assessed page. They help expose ambiguous identity, unclear offerings, missing decision details, or inconsistent evidence. Results are calibrated and shown only for models that completed the run.

These are directional model-comprehension tests. They do not mean a signed-in consumer application independently discovered the page, and they are not rankings inside ChatGPT, Gemini, Claude, Meta AI, Grok, Amazon, or another consumer product.

AI Visibility

Question answered: Across configured checks, which brands are mentioned for the questions being tracked, and how does that evidence change over time?

AI Visibility records supported prompt runs, detected brand mentions, competitors, and citations when a provider returns them. The workspace keeps the run history so customers can inspect changes rather than treating one answer as permanent truth.

• Provider access methods and citation availability vary.
• AI answers can vary between runs, locations, models, and dates.
• A mention is not automatically a recommendation, citation, click, or conversion.
• Results should be treated as directional evidence unless the product identifies a grounded source for the specific run.

AI Journeys

Question answered: Across a multi-step customer decision, where does a business enter, remain in, or leave consideration?

AI Journeys follows a decision from discovery through comparison and choice. Supported grounded Journey surfaces use current search-connected provider capabilities and retain citations from each qualifying run. Multi-platform execution is an explicit plan choice because coverage and provider costs vary.

Repeated runs help show variance. An Exact Rerun replays the saved question sequence so the before-and-after comparison is not silently changed by generating a new journey. A Journey result is evidence from the tested surfaces and run conditions—not a universal prediction of every customer's private conversation.

Agent Replay

Question answered: Can a bounded AI agent use a verified website to reach an intended step safely?

Agent Replay is separate from conversational AI Journeys. It uses a browser-action test on a customer-verified website, follows explicit time and action limits, and stops before purchases, submissions, registrations, or other final actions. It does not solve CAPTCHAs, bypass access controls, or leave the verified-site boundary.

Results distinguish a completed safe path, a fixable dead end, a blocked attempt, and an unable-to-test outcome. Unable to test does not mean the website has a problem. It means BoostGeo could not collect enough safe evidence from that run.

Fix Center and exact verification

Question answered: Did the specific condition linked to a finding change after the customer acted?

Fix Center keeps the evidence, proposed action, and verification target connected. Where supported, a customer can approve a controlled publication, retain a versioned change record, verify the public result, and roll back safely. A completed task alone is not counted as proof that the issue improved; BoostGeo distinguishes work performed from a passed recheck.

Verified Facts

Question answered: Which business facts has an authorized customer reviewed and approved for consistent machine-readable publication?

Verified Facts creates a controlled source for customer-approved information and supported publication formats. Customer approval establishes ownership and intent; it does not, by itself, make a fact independently verified by BoostGeo. Publishing or notifying a supported index does not guarantee indexing, citation, ranking, or use by a consumer AI product.

Outcome and attribution evidence

Question answered: What observable visibility, referral, traffic, or conversion change followed during a clearly stated period?

When an analytics connection is available, BoostGeo can compare clearly labeled periods and retain the relationship between a finding, intervention, recheck, and later observation. This helps customers investigate impact while preserving the difference between sequence and causation.

BoostGeo reports observed change and correlation, not guaranteed causation. Other marketing activity, seasonality, provider changes, and outside events can affect results.

Why there is no single universal score

Website readiness, model comprehension, grounded visibility, journey inclusion, browser-task completion, and business outcomes are different questions. Combining them into one opaque number would hide those differences and could imply more certainty than the evidence supports. BoostGeo instead connects the measurements through one operating loop: measure, improve, publish, verify, and prove.

What BoostGeo does not promise

• Guaranteed rankings, mentions, citations, recommendations, traffic, or revenue.
• Exact replication of signed-in or personalized consumer AI experiences.
• Access to private model training data, internal context, or proprietary ranking systems.
• That a crawler-access rule, structured-data change, machine-readable file, or published fact will cause an AI platform to use the content.
• That correlation after a fix proves the fix caused the observed business outcome.

Questions about a result

If a result appears inconsistent with the evidence shown, use the contact form. Include the affected page and run date if available. For information about the BoostGeo Score Bot, see the crawler identity and access policy.