Methodology

How Bromley scores your visibility in AI answers

Every grade Bromley reports is traceable to a check we actually ran. No black box, no vanity metric — this page is the whole framework: what we measure, how the number is built, which AI surfaces we test, and what we deliberately don't claim to know.

Grounded in Google's own guidance, not folklore

Google has been explicit: answer-engine and generative-engine optimization are not separate disciplines — they are SEO applied to generative surfaces. Bromley follows Google's published framework rather than the proliferation of unverified tactics (llms.txt files, content chunking, manufactured mentions) that the official guidance explicitly de-prioritizes. Three pillars, in the order an AI answer actually depends on them:

Pillar i

Foundational SEO

Technical structure, indexability, page experience, JavaScript rendering, duplicate content — the substrate every AI feature draws from via retrieval.

Pillar ii

People-First Content

Unique point of view, first-hand experience, non-commodity insight, clear structure and named authors — weighed against Google's helpful-content guidance.

Pillar iii

Generative Surface Testing

Live probing of the AI answer surfaces themselves for whether your site is actually cited on the queries that matter to you.

How the score is built

Each pillar's score is simply the share of its checks that passed, mapped to a letter on a standard scale. The headline number is the equal-weighted average of the pillars we could actually measure this run — no pillar is weighted more heavily to flatter a result.

A 90–100 B 80–89 C 70–79 D 60–69 F below 60

The honest part A pillar that didn't produce enough signal this run is left out of the average rather than guessed at. Your grade reflects only what we measured — so it never invents confidence we don't have, and it stays comparable run-to-run as coverage grows.

The AI surfaces we test

Pillar iii is measured by asking the real answer engines your tracked queries and reading back whether your site is cited. Today that's:

ChatGPT Perplexity Google AI Overviews

One capture, kept as a receipt For each tracked query on each surface we take one live capture and store the verbatim answer we received — the exact text, its sources, and whether your brand appears. That captured answer is the receipt behind every citation claim in your report; you can open it yourself. We add surfaces as coverage expands, and we don't report a number we didn't capture.

What we do not audit

Naming the boundary is part of being honest about the grade:

Paid placement. Ads and sponsored answers aren't earned visibility, so they don't count toward your score.
Private or logged-in surfaces. We test the public answer a prospect sees, not personalized results tied to someone's account or history.
Manufactured mentions. We don't buy, plant, or fabricate citations — the audit measures whether the open web earns them.
Unverified tactics. Techniques Google's guidance de-prioritizes (llms.txt, keyword chunking) aren't scored as if they were established ranking factors.

Want to see this run against your own site? Bromley captures the answers, grades each pillar, and shows you every receipt.

Run a free audit →