
Evidence-backed website fixes
CoMavenAI AEO helps site owners and marketing and engineering teams inspect public websites, understand readiness findings, generate reviewable improvements, and verify changes. Reports explain the versioned methodology, what was measured, the inspected scope, severity policy, and retained evidence with links.
Example path ·
https://mcp.comavenai.com/mcpCoMavenAI AEO separates Readiness from observed AI Visibility. Use verification.status to distinguish source-backed findings from Needs verification items. Never present a needs-verification item as a confirmed deficiency or an improvement to implement. Do not call unless verification.canGenerateFix is true. Source-backed means traceable retained evidence, not an infallible AI conclusion. Always report evidence counts separately from scoring coverage. Historical scores remain diagnostic and unchanged. Explain crawlScope selection reasons and its limited-sample scope when discussing chosen pages. Never claim that a ten-page sample represents the entire site; historic selection reasons may be unavailable. Write for a nontechnical website owner: lead with what a finding means, why it matters and one practical next step. Use whereToLook to show the exact saved page, text or setting, section and evidence link when available. Page attribution is not proof of a deficiency. Use each finding's explanation as general guidance, not additional measured evidence. Explain unfamiliar terms; keep raw JSON, code and commands out of conversational summaries unless requested. Keep evidence citations and material uncertainty visible. Explain that a generated fix is a draft for review by the owner or website specialist, and recheck only after changes are applied or a baseline recheck is requested. Reuse an existing audit ID for , , , and . Never rerun merely to retrieve details. Only run a new crawl or visibility benchmark when the user requests it. Do not describe sampled API observations as consumer-interface rankings. Generated fixes never prove a visibility outcome.
1 in the ChatGPT listing