How to Build an AEO Audit Report in 90 Minutes (The Template Agencies Charge $2,500 For)

Your prospect on Wednesday's discovery call is going to ask for an AEO audit before they sign. They want a deliverable that looks expensive, ships fast, and tells them whether the SEO retainer they have been paying for is still doing anything in an AI-search world. Most agencies sell this audit as a one-off project for $2,500 and spend three weeks producing something that should take 90 minutes.

This is the template I run. Five sections, ninety minutes, one productized deliverable. Every signal in the audit is sourced from published research. None of it requires custom code. The only platform line item is whatever AEO tracking tool you already use; the rest is method, not magic.

The market context for offering this as a deliverable is unambiguous. 77% of brands are completely invisible in AI platform answers per Loamly's 2026 benchmark of 2,089 brands. The 23% that are visible convert AI-sourced traffic at three times the rate of Google Search. 94% of B2B buyers use LLMs during their buying journey per the 6sense Buyer Experience Report. And 94% of CMOs plan to increase AEO/GEO investment in 2026 per Conductor's State of AEO/GEO 2026 report. The audit is the wedge into that increased budget.

Pricing the audit at $2,500 sits in the defensible middle. Per Eagles Media's published agency rate analysis and Relixir's 2025 AEO pricing breakdown, project-based AEO audits typically run $5,000-$15,000 with full retainers between $1,500 and $10,000/month. A productized $2,500 audit is the upper bound of what freelance specialists charge on Upwork and the lower bound of what agency retainers anchor against on a first engagement. Above that you owe the prospect customization. Below that you give the work away.

The 90-minute structure (5 sections × ~18 minutes each)

The version I run for client engagements. Each section produces one defensible artifact. The total of all five is the audit deliverable.

Section 1 — Visibility baseline (15 minutes)

Open ChatGPT, Perplexity, Gemini, Claude, and Google AI Overviews in incognito windows. Run 3-5 queries that map to the client's top buyer intents — not keyword phrases. Run each query three times. Note the spread. Per Ahrefs' published practitioner method, 3 runs across 5 engines for 5 queries is 75 data points, which is enough signal to detect mention consistency without falling into single-snapshot bias.

Capture two distinct measurements per engine. Mentions: how many times the brand name appears anywhere in the response text. Citations: how many times a URL on the brand domain is cited as a source. Per Ahrefs, these are different signals and most audits conflate them. A brand can be mentioned in narrative text without ever being cited as a source URL, and vice versa. Both matter. Track both.

Section 2 — Content + E-E-A-T (20 minutes)

Pick the ten URLs most likely to be cited. Run a structured E-E-A-T checklist on each: visible author byline, published and updated dates, content chunked into H2/H3 sections, FAQ blocks, expert credentials, internal links to authority pages, user reviews or social proof. Score each as pass, partial, or fail.

The key signal to capture: is the page chunked into 50-150 word self-contained sections with Q&A headings? AI engines extract those chunks more readily than they extract long-form prose. The agency-side observation here is that most clients hand you exactly the wrong content shape — long flowing paragraphs that read well to a human but extract poorly to a model.

Section 3 — Schema + structured data (15 minutes)

Audit existing schema. Flag generic schema as a regression rather than a baseline. Per Growth Marshal's analysis, pages with attribute-rich Product, Review, or FAQ schema were cited at 61.7%; pages with no schema were cited at 59.8%; pages with generic Article or Organization schema were cited at 41.6%. Generic schema underperforms doing nothing. The audit recommendation should never be "add schema." It should be "add attribute-rich schema, or strip the generic schema you have."

Per Frase's research, pages with FAQPage markup are 3.2× more likely to appear in Google AI Overviews. FAQ schema is the highest-ROI single addition for most clients. Identify the top three pages where FAQ schema is missing and would map to real shopper questions.

Section 4 — Citation + Reddit (20 minutes)

This is the section most agencies skip and the section that produces the most retainer-defending insight. Map where the brand currently is and is not cited.

Per Discovered Labs' citation analysis, Reddit accounts for 46.7% of Perplexity's top 10 citations. Per Semrush's analysis of 248,000 cited Reddit posts, 80% of cited Reddit content has fewer than 20 upvotes — AI engines pull from comment chains and niche subreddits, not from megaviral threads. The implication: the agency can credibly compete in Reddit citations through educational comments rather than through top-of-feed posts.

Per Ahrefs' analysis of 75,000 brands, YouTube mentions correlate 0.737 with AI visibility — the strongest single signal across all platforms. A client with no YouTube presence is invisible to a meaningful share of AI source weighting. Capture the gap.

Section 5 — Technical + measurement (20 minutes)

Three checks here. Crawl access for ChatGPT and Gemini user agents (most clients accidentally block one or both via robots.txt). Core Web Vitals on the top 10 citeable pages (slow pages get cited less). GA4 attribution coverage, which is the single most-broken layer in the client's stack.

Per Coalition Technologies' analysis, default GA4 setups correctly classify only 0.5% of ChatGPT traffic; the rest sits unattributed in the Direct bucket. AI visitors who do click through spend 68% more time on the site per the same Coalition study. The audit should always include a GA4 channel-group fix as a quick win — it converts invisible AI conversions into reportable ones for the next monthly review.

Skip llms.txt as a priority recommendation. Per Search Engine Journal's coverage of SE Ranking's 300,000-domain study, only 10.13% of measured domains have an llms.txt file and there is no statistical correlation with AI citation frequency. Mention it as an emerging signal to monitor; do not stake the audit recommendation on it.

The five red flags an audit should never ship with

The bad-audit warning signs separating professional work from vendor marketing.

Single-engine testing. A score from ChatGPT alone or Google AI Overviews alone misses 80% of the picture. Per Loamly's cross-engine analysis, ChatGPT and Gemini cite the same brands only 19% of the time. A multi-engine baseline is non-negotiable.

No confidence intervals. AI responses vary run to run. A claim that "the brand has zero mentions" after a single test is unreliable. Three runs per query, with the spread captured, is the minimum defensible methodology.

Generic schema recommendations. Telling a client to "add Article schema" when they already have generic Article schema is actively harmful per Growth Marshal's data — generic schema cites worse than no schema at all. The recommendation has to be attribute-rich Product, Review, FAQ, or LocalBusiness schema, populated with real fields.

llms.txt as the lead recommendation. The 300K-domain study found zero correlation. If the audit lead is "implement llms.txt," the audit is selling vendor narrative rather than measured impact.

No attribution methodology. Reports that claim "you got X AI traffic" without explaining how AI traffic was identified are misleading because GA4 captures only 0.5% by default. The audit should always include the GA4 fix as part of the deliverable.

What "valid AEO data" looks like (the GenPicked Research Team reference)

Before you ship the audit, hold it to the standard of validity. The Conductor State of AEO/GEO 2026 report documents that 97% of CMOs reported positive AEO impact in 2025, but the methodology that underpins those reported gains varies wildly. Per Conductor's CMO survey, the leaders who report the largest gains are also the ones whose tracking methodology includes confidence intervals, model splits, and sycophancy diagnostics — not single-number aggregate scores.

The GenPicked Research Team (2026) Fitness Wearables Study shipped a worked example: a Bradley-Terry maximum-likelihood ranking across four AI models with 95% confidence intervals. Oura ranked 1.82 [1.71, 1.94], Whoop 1.44 [1.29, 1.58], Garmin 0.92 [0.78, 1.07] — with the intervals showing Oura statistically separated from Whoop, but Apple Watch and Fitbit at the bottom in a statistical tie. That is what an audit defensible at the boardroom level looks like. A score without an interval is a claim without evidence.

For agency-side audits, the practical translation is more modest: report mention and citation rates per query, track the spread across three runs, and never collapse five engines into a single averaged number. That is enough rigor to defend the deliverable through a CFO conversation, which is the conversation most $2,500 audits will eventually face.

Pricing the audit and what comes next

The $2,500 audit is the wedge, not the retainer. Per Seer Interactive's September 2025 analysis of 3,119 informational queries, brands cited in AI Overviews receive 35% more organic clicks and 91% more paid clicks than uncited competitors. That delta is the retainer math — the prospect either acts on the audit findings within 30 days or watches the gap to their cited competitors widen. Most agencies should price the audit slightly above their cost basis, ship it fast, and use the findings as the brief for a 90-day implementation engagement at $3,000-$5,000/month.

The platform tooling layer matters at scale. A single $2,500 audit can be run by hand. Twenty audits per month cannot. The structural shift happens around the fifth concurrent client — that is when manual five-engine tracking, schema audits, and Reddit citation mapping start consuming more agency hours than the retainer covers. The audit template productizes the deliverable; the platform productizes the delivery.