VisibilityTrace covers a category — AI visibility tools, LLM rank trackers, GEO and AEO platforms — where vendor marketing, affiliate listicles, and independent research are mixed together at every level of search. The methodology described on this page exists to keep those three layers separate on every page of the site. Each page declares, in its header, what kind of research produced its claims.
The five evidence labels
Every page on VisibilityTrace carries exactly one evidence label. The label sets the reader's expectations about what was done to produce the page and what was not done. The five labels in use are listed below.
Deep Research buyer guide
A buyer guide drafted from a structured Deep Research pass over public documentation, pricing pages, changelogs, vendor blog posts, third-party empirical studies, and academic preprints. The research prompt and the source list are operator-reviewed before drafting begins. Each factual claim in the final page links to its primary source with an observed date.
Public information review
A profile of a single vendor built from the vendor's own public pages — pricing page, documentation, terms, public roadmap. An account may have been created to confirm what is visible to a logged-in prospect, but no usage testing is performed. The label uses the phrase account-created-not-tested to make the scope explicit. Source URLs are operator-approved before publication.
Public data comparison
A side-by-side comparison built from quantitative data that is publicly available for each compared vendor — published pricing, declared engine coverage, declared sampling frequencies, declared panel sizes. Cells where a vendor does not publicly disclose the relevant figure are marked as such, not interpolated.
Methodology-based comparison
A comparison that applies documented evaluation criteria — for example, "methodology transparency," "API-versus-UI disclosure," "freshness of third-party validation" — to publicly available information about each tool. The criteria are stated up front. The page does not rank tools on any criterion that has not been declared at the top of the page.
Policy and trust page
A page about VisibilityTrace itself — affiliate disclosure, update policy, about page. Drafted from operational facts about how the site is run. Low editorial risk; reviewed by the operator before publication.
The Deep Research process
Deep Research is the heaviest production path on VisibilityTrace and the only path used for category buyer guides. It runs in five stages.
- Scope and seed sources. The operator confirms the category boundary (for example: "LLM rank trackers, not generic SEO suites"), the candidate vendor list, and a starter set of seed sources — academic preprints, third-party empirical studies, vendor methodology pages.
- Research prompt. A research prompt is drafted that names the evaluation criteria explicitly: what to look for, what to compare, what conflicts to surface. The operator reviews the prompt before any research run starts.
- Parallel research runs. The same research prompt is run across two independent research systems. Running the same prompt twice in parallel surfaces disagreements between the two outputs and reduces the risk of a single research system fabricating a citation.
- Source verification. Every primary citation that the research outputs will rely on is opened, read, and the relevant figure or quote is confirmed. The observed date is recorded for each source URL.
- Drafting and claim review. The page is drafted from the verified source set. Each factual claim links to its source with the observed date. The operator reviews claim-by-claim before publication.
The Deep Research archive for each buyer guide is kept in the repository alongside the page: the two parallel research outputs, the prompt that produced them, and a short note describing where they agreed and disagreed.
The public information review process
A public information review is a lighter production path used for single-vendor profile pages. The constraint is that no claim leaves the vendor's own published surface area unless that claim is also confirmed by an independent public source.
- Source pack. The operator approves a list of vendor public pages that the review will cite: pricing page, documentation index, terms of service, public roadmap or changelog, public blog posts of substantive content. Pages that require an account to view are not used as primary sources for a public information review.
- Drafting. The review summarises what the vendor publicly claims about features, pricing, plan tiers, engine coverage, and policy. Each claim is tied to the specific page it came from, with an observed date.
- Marketing-claim separation. Where a vendor publishes a claim that is not independently verifiable from a public source — for example, "the most accurate LLM rank tracker" — the claim is presented as a vendor statement and flagged as such, not promoted to a factual assertion by the page.
- Operator approval. The full claim set is reviewed before publication. Affiliate CTAs on the page are not enabled until the review is approved.
BOSS Pipeline operational evidence
Several pages on VisibilityTrace — the AI crawler readiness checklist, the "what AI bots see" explainer, parts of this methodology page — are informed by operational evidence gathered from running the BOSS Pipeline, a fleet of static affiliate-monetised content sites operated by the same operator. That evidence is used under one constraint: no specific site in the fleet is identified by name, domain, or any other recognisable handle.
The fleet provides three categories of operational input:
- Crawler-access patterns. Server logs from a multi-site deployment show which AI crawler user agents request which paths, at what cadence, and what response codes they receive. This is the basis for the AI crawler readiness checklist's "common failure modes" section.
- Static-versus-dynamic rendering outcomes. The fleet has gone through migrations between SPA-style rendering and static HTML generation. The "what AI bots see" page draws on the visible differences between those two regimes when probed with crawler user agents.
- Affiliate program operations. The build-time affiliate-reference resolver pattern described on the affiliate disclosure page is the same pattern used across the fleet. The pattern was developed because raw affiliate URLs in committed source code are a recurring cause of leaked references and broken partner relationships.
BOSS Pipeline operational evidence is treated as a primary source on the same footing as any external public document — it is cited (without naming sites) and dated where dating it is meaningful.
What VisibilityTrace does not claim
- No hands-on benchmark testing of every tool covered. Where account-level inspection has happened, the page says so explicitly and labels itself account-created-not-tested.
- No guarantee of AI citation outcomes or AI visibility improvements from using any tool covered.
- No official partnership with Google, OpenAI, Anthropic, Perplexity, or any other AI vendor.
- No real-time accuracy of vendor pricing or terms. Every cited figure carries an observed date and a link to the source page.
- No "best overall" claim that is not pinned to specific evaluation criteria stated up front on the page.
Enforcement
The discipline described above is enforced by automated checks that run before any page can be built into the live site. The checks include: a forbidden-wording scan that fails the build on the no-go phrases listed above, a schema validator that restricts JSON-LD types per page class, a static-body probe that confirms each page renders its H1 and main text without JavaScript execution, a canonical audit, and a sweep against any raw affiliate URLs in committed source.