Our position: Treat AEO audits like evidence reviews, not creative brainstorms. The agency should prove the visibility problem, classify the answer behavior, and separate known facts from strategic guesses. Recommendations come later, after the audit has made the client’s answer-engine gap visible.
What you should leave with
- A useful AEO audit template starts with scope control: market, location, platform set, prompt families, competitors, and classification rules.
- The audit should capture answerability, crawlability, citation ownership, entity consistency, source coverage, factual accuracy, and retest readiness.
- Client deliverables should convert raw AI answers into business gaps, prompt evidence, source patterns, wrong facts, and a sequenced fix plan.
- Agencies should avoid fake authority scores, unexplained screenshots, paid placement shortcuts, and any claim that guarantees AI recommendations.
What are the required sections in an agency AEO audit template?
The required sections are scope, prompt set, AI answer capture, source and citation review, website evidence review, entity and schema review, factual accuracy review, opportunity classification, and delivery recommendations. Each section needs a field owner, an evidence requirement, and a QA gate so the audit can be repeated by more than one analyst.
AEO audits fail when the template lets the analyst jump straight from one odd AI response to a broad recommendation. The template should slow the work down. It should ask what was tested, what appeared, what source supported it, whether the client controlled that source, and whether the answer was accurate enough to act on.
Use the free audit to generate the first client evidence row before you scale the process. Once the same fields work across several clients, move the workflow into agency white-label pricing so sales, delivery, and reporting use the same evidence structure.
| Section | Required Fields | Review Gate |
|---|---|---|
| Scope lock | Market, location, audience, competitors, platforms, date range | Client or strategist confirms the test universe |
| Prompt inventory | Prompt family, buyer stage, intent type, exact prompt, variants | No testing begins until prompts are approved |
| Answer capture | Platform, answer summary, recommendation role, citations, screenshots or export | Evidence row includes raw answer location |
| Source review | Cited domains, client-owned pages, third-party pages, missing sources | Every cited source is categorized |
| Accuracy review | Correct facts, wrong facts, outdated facts, unsupported claims | Risky claims are separated from visibility issues |
| Fix plan | Issue, impact, owner, content or technical action, retest prompt | Every recommendation maps to evidence |
Evidence used in this section
How should agencies scope the audit before testing?
Agencies should freeze the audit scope before testing by defining the market, location, buyer type, competitor set, platforms, prompt families, and classification policy. This protects the audit from becoming anecdotal. Without a locked scope, every strange answer feels important, and the client cannot tell whether results changed because the market changed or the testing changed.
The scope section is the audit’s control system. It should say which answer engines are being tested, which use cases matter, which competitors count, and which prompts are in or out. A SaaS client, a local service business, and a regulated professional service firm should not share the same prompt universe.
Borrow the govern, map, measure, and manage mindset from AI risk management: define the system, map the context, measure observed behavior, and manage the resulting risks. That discipline keeps the agency from overclaiming what the audit can prove.
- STEP 1
Freeze the market
Freeze the market: category, geography, buyer segment, and language.
- STEP 2
Freeze the competitor set
Freeze the competitor set: direct competitors, substitutes, marketplaces, and publishers.
- STEP 3
Freeze the platforms
Freeze the platforms: ChatGPT search, Google AI features, Perplexity, Gemini, or the agreed testing set.
- STEP 4
Freeze prompt families
Freeze prompt families: comparison, recommendation, problem-aware, alternative, pricing, implementation, and risk prompts.
- STEP 5
Freeze classification policy
Freeze classification policy: how the team labels recommendation role, citations, factual accuracy, and confidence.
Method boundary: Do not add new competitors or prompts midway through the audit unless they are logged as a second test wave. Otherwise, the evidence table stops being comparable.
Which AEO checks belong in the template?
The core checks are answerability, crawlability, visible evidence, schema accuracy, entity consistency, source coverage, factual accuracy, and retest readiness. These checks connect website quality, third-party authority, and AI answer behavior. They also stop the agency from treating AEO as only content production or only technical SEO.
Answerability asks whether the client has a clear, useful answer to the questions buyers actually ask. Crawlability asks whether important pages can be discovered. Visible evidence asks whether claims, comparisons, pricing, policies, proof, and expertise are present on the page in a form a user can verify.
Schema accuracy matters, but it is not a shortcut. Structured data should describe what is visibly available on the page. If the FAQ block, article details, or how-to steps are not present for users, schema should not pretend they are.
- Answerability: Does the client answer the prompt directly and usefully?
- Crawlability: Can important pages be accessed and indexed by relevant crawlers?
- Visible evidence: Are claims supported with proof, examples, data, or clear explanation?
- Schema accuracy: Does structured data match the visible page content?
- Entity consistency: Are names, services, locations, founders, and categories consistent across sources?
- Source coverage: Are credible third-party references present where answer engines are likely to look?
- Factual accuracy: Are AI answers repeating wrong, stale, or unsupported facts?
- Retest readiness: Is there a specific prompt and page set to test again after fixes?
Evidence used in this section
How should the agency classify AI answer results?
The agency should classify each AI answer by recommendation role, citation ownership, competitor advantage, factual accuracy, and confidence. This turns messy answer captures into a usable decision table. The goal is to identify whether the client is absent, mentioned weakly, cited without being recommended, misrepresented, or beaten by better-supported competitors.
Classification is where the audit becomes useful. A raw answer may mention the client, but that does not mean the client won the prompt. The answer might recommend a competitor, cite a marketplace, summarize outdated information, or include the client only as an afterthought.
Keep confidence humble. AI answers vary by platform, time, personalization, and retrieval path. The audit can show observed patterns and evidence gaps. It should not promise permanent placement, fixed rankings, or guaranteed inclusion in AI-generated recommendations.
| Field | Labels | Why It Matters |
|---|---|---|
| Recommendation role | Primary recommendation, secondary mention, neutral mention, absent, negative mention | Shows whether visibility is commercially meaningful |
| Citation ownership | Client-owned, partner, publisher, marketplace, competitor-owned, uncited | Shows who controls the evidence being used |
| Competitor advantage | Better proof, broader source coverage, clearer category fit, stronger reviews, pricing clarity | Explains why another brand may be preferred |
| Factual accuracy | Accurate, partially accurate, outdated, unsupported, wrong | Separates visibility loss from brand risk |
| Confidence | High pattern, medium pattern, single observation, needs retest | Prevents overinterpretation of one response |
Evidence used in this section
How does the audit become a client deliverable?
The audit becomes a client deliverable when the agency converts evidence into business gaps, prompt examples, source patterns, wrong-fact risks, and a sequenced fix plan. The client should be able to see what was tested, what the AI answer did, why it matters, and what action comes next.
A strong deliverable is not a folder of screenshots. It is a narrative built from rows of evidence. Start with the business gap: the client is missing from high-intent recommendation prompts, losing citations to directories, or being described with outdated facts. Then show the prompt evidence that proves the pattern.
The fix plan should be sequenced. First repair facts and crawlability problems. Then improve owned evidence pages. Then close source gaps with credible third-party proof. Then retest the original prompt set. That order helps clients understand why the agency is not simply publishing more content.
- STEP 1
Summarize the commercial visibility gap in
Summarize the commercial visibility gap in plain language.
- STEP 2
Show representative prompt evidence, not every
Show representative prompt evidence, not every raw capture.
- STEP 3
Group citations into source patterns
Group citations into source patterns: owned, earned, directory, review, publisher, competitor, or missing.
- STEP 4
List wrong or stale facts separately
List wrong or stale facts separately as brand-risk items.
- STEP 5
Sequence fixes by urgency, owner, effort,
Sequence fixes by urgency, owner, effort, and retest method.
Method boundary: Raw AI answers can be included in an appendix, but the main report should explain patterns. Clients buy judgment, not a pile of transcripts.
What should agencies leave out?
Agencies should leave out fake authority scores, unexplained screenshots, undisclosed paid placement requests, unverifiable benchmark claims, and any promise that the client will be recommended by AI systems. These items make the audit look more confident while making it less trustworthy.
The worst AEO audit templates imitate SEO scorecards without proving what the scores mean. A made-up AI authority score may look neat in a dashboard, but it often hides the actual question: which prompts did the client lose, which sources shaped the answer, and which facts need repair?
Also leave out recommendations that imply undisclosed manipulation. If the plan involves reviews, endorsements, partnerships, sponsorships, or paid placements, the audit should respect disclosure rules and avoid presenting paid influence as organic authority.
- Do not invent a single authority score unless the method is transparent and useful.
- Do not paste screenshots without prompt, platform, date, and interpretation.
- Do not recommend paid placement without disclosure and claim boundaries.
- Do not guarantee rankings, citations, recommendations, or answer inclusion.
- Do not use schema to mark up content that is not visible to users.
- Do not recommend content solely for machines while ignoring human usefulness.
Method boundary: AEO is measurable, but it is not fully controllable. The audit should create better evidence and better retesting, not pretend the agency can command every generated answer.
Evidence used in this section
Questions that change the decision
Frequently asked questions
How is an AEO audit template different from an SEO audit template?
An SEO audit template usually focuses on rankings, indexing, technical issues, content gaps, and backlinks. An AEO audit template focuses on how answer engines respond to buyer prompts, which sources they cite, whether the client is recommended, and whether generated answers are accurate.
Can an agency white-label this AEO audit checklist template?
Yes. The template is suitable for white-label use if the agency keeps the evidence fields, source categories, claim boundaries, and QA gates intact. Branding can change, but the audit method should stay consistent so results remain repeatable.
How many prompts should be included in the first AEO audit?
Start with 20 to 40 prompts across buyer stages and prompt families. That is usually enough to reveal patterns without creating noise. Larger clients may need more, but prompt quality and classification discipline matter more than raw prompt volume.
Should raw AI answers be included in the client report?
Include selected raw answers as evidence and keep the full capture set in an appendix or working file. The main report should summarize patterns, risks, and recommended actions. Clients need interpretation more than long transcripts.
What should the post-audit deliverable include?
The post-audit deliverable should include the business gap, prompt evidence, source and citation patterns, factual accuracy issues, priority fixes, owners, and retest prompts. It should also state that AI visibility can be improved and measured, but not guaranteed.
Primary sources and research
Platform documentation supports factual statements. Where we describe an audit method or prioritization rule, that is AnswerMentions' operating judgment and is labeled as such.
- [1]Google Search Central: creating helpful, reliable contentGoogle recommends original information, substantial analysis, clear sourcing, and content that leaves a visitor feeling they learned enough to achieve the goal.
- [2]Google Search Central: structured data policiesGoogle requires structured data to match visible content and makes clear that valid markup does not guarantee a search feature or recommendation.
- [3]NIST: AI Risk Management FrameworkNIST frames AI risk work around governance, mapping, measurement, and management, a useful model for separating observations from decisions.
- [4]FTC: advertising and marketing basicsThe FTC states that advertising claims must be truthful, non-deceptive, and supported by evidence when appropriate.
- [5]Schema.org: FAQPageFAQPage defines machine-readable questions and accepted answers; the visible content remains the substance that users and systems evaluate.
- [6]Schema.org: HowToHowTo schema describes visible step-by-step instructions when the page actually contains those steps.
- [7]FTC: reviews and endorsements guidanceFTC guidance treats reviews and endorsements as claims that need honest representation and appropriate disclosure, not as raw material to manufacture social proof.
- [8]OpenAI Help: accuracy and citationsOpenAI warns that ChatGPT can produce incorrect facts and fabricated references, so consequential claims should be checked against reliable sources.