← Back to Verdict∞

How the Verdict∞ Audit Actually Works

Published 21 August 2026 · By Stephen Steddy, Co-Founder, Sage8.AI

Most website “audits” are a template with your company name swapped in. This one isn't. Every Verdict∞ report is generated fresh, per site, by an AI agent that reads your live pages and produces structured, schema-validated findings — not freeform prose a human then has to interpret. Here's exactly how that happens.

1. Context intake

Before anything is fetched, we take your URL, industry, target market, target audience, business goal, platform age, and any competitors you already know about. This context shapes how the audit is scored — a two-person consultancy and an e-commerce brand aren't judged against the same bar.

2. Live site fetch — not a cached crawl database

The audit agent fetches your actual pages directly, in real time, via a dedicated retrieval tool — reading structure, content, and code as they exist right now. It isn't drawing on a stale third-party index. If a page can't be reached, that gets flagged as an access issue in your report rather than scored as if it were fine.

3. Nine-dimension scoring

Every dimension — SEO, AEO, GEO, LSO, LLM Citation Potential, Brand, UX, Grammar, and Links — is scored against the evidence gathered in step 2. What gets returned depends on your tier:

The findings are submitted as structured, schema-validated data — not freeform text — and rejected and re-generated automatically if a required section is missing or malformed before your PDF or .docx is ever built. That's what keeps quality consistent across every audit, rather than varying with how thorough a given run happened to be.

4. Competitor benchmarking

For the Findings Report and Full Report tiers, the agent runs live web search to find real competitors in your market — either the ones you named, or ones it identifies — and benchmarks your site against them directly. It doesn't generate a plausible-sounding generic competitor list; if a competitor genuinely can't be researched, that's disclosed rather than filled in.

5. The Free Re-audit delta

When you redeem your Free Re-audit, the engine is given your original report as a scoring baseline and audits your site fresh — it doesn't assume you've improved. The new report then shows what actually changed against that baseline, so the comparison reflects reality rather than good faith.

The Live AI Engine Check

Most of what a Verdict8 report measures — AEO, GEO, LLM Citation Potential — is an inferred score: how well-structured your site is for an AI engine to find and cite. To go one step further, we also run a live check against real AI engines: a set of realistic buyer-intent prompts (the kind an actual prospective customer would type — not “what is [your business]”) are sent to Claude, and — where configured — ChatGPT and Perplexity, and we check whether your domain is actually mentioned or cited in the response. We report this as a directional, real-time snapshot alongside the inferred score, not a certified ranking — because it is one run, on one day, and AI answer engines are non-deterministic by nature.

Verification over flattery

Verdict∞ is a diagnostic and advisory service, not a guarantee of results — we don't promise a specific ranking, traffic increase, or business outcome. Findings that can be checked directly against your live site are checked, not guessed. If we can't verify something, the report says so instead of inventing a score. That's also why Findings Report and Full Report both go through a human review step before they reach you — a person checks the agent's work before you see it.

See the full framework on the homepage, or read our frequently asked questions for pricing and process details.