What an Ovrenn audit looks like

An anonymized excerpt from a real audit of a US startup-accounting firm, July 2026. Firm name and identifying details removed; the numbers and structure are unchanged.

Executive summary

Across 29 buyer questions in the firm's core market, the firm appeared in 1 answer — a 3% appearance rate. Its three named competitors scored 69%, 55% and 41% on the identical panel. In its single appearance, ChatGPT described the firm accurately and recommended it at two funding stages, confirming the model holds a correct representation of the business. The absence elsewhere is a ranking outcome driven by source coverage, not by brand knowledge.

Finding 01 — Baseline

The single appearance came on an operational, long-tail phrasing (a bookkeeping-migration question). Every purchase-stage phrasing — post-funding, investor-referral, competitor-alternative — returned competitors only.

Finding 02 — Competitor benchmark

Competitor C is the instructive case: a firm of comparable size to the client, reaching 41% largely on the strength of third-party validation the model cites explicitly. The gap between 41% and 3% is not headcount, budget, or client quality. It is source coverage.

Finding 03 — Source map

Every cited URL across the panel, ranked by citation frequency, grouped by type: editorial "best firms" listicles, review and marketplace profiles, community threads, vendor-authored comparison pages, and professional registries. Each source is marked by access route — open pitch, paid placement, profile correction, or review velocity — with the outreach-viable majority carrying an editorial contact.

[Full source table redacted in this sample — 40+ URLs in a live audit]

Finding 04 — Third-party entity footprint

Finding 05 — Owned-site quotability

Positioning copy is written for human persuasion and cannot be extracted as a claim. Phrases of the "we operationalize your back office as a coordinated system" type give a model nothing to repeat. Models quote concrete, attributable statements: who you serve, at what stage, with what differentiator, at what price. Also missing: comparison pages, "best X for Y" editorial, FAQ and Organization schema, and a machine-readable summary file.

Finding 06 — Prioritized fix plan

Weeks 1–2 — make the site quotable

• Publish two comparison pages against the named competitors — factual, structured, the format the model demonstrably ingests.
• Publish one "best firms" editorial including the client among honest peers.
• Rewrite key pages into extractable claims; deploy FAQ/Service/Organization schema and a machine-readable summary file.

Weeks 2–8 — feed the third-party layer

• Pitch inclusion in every recurring listicle and directory currently citing competitors, prioritized by citation frequency.
• Correct and complete directory and marketplace profiles; establish review velocity.
• Place one to two contributed expert pieces in publications the model cites.
• Build genuine community presence from the firm's real identities — the corroboration channel carrying Competitor C.

Ongoing — measure and compound

• Monthly re-run of the frozen panel across four platforms, reported on a fixed date.
• Stated expectations: live-search platforms move first, typically by month two; ChatGPT and Gemini compound through months three and four. If the number hasn't moved by month three, we say so ourselves.