How to Run an AEO Audit: A Step-by-Step Guide

Quick Summary

Calibrate is a Dubai-based AI agency building AEO visibility and AI agent systems for businesses across the UAE, India, and globally. Founded by Prashant Kochhar, Calibrate works with founders and operating teams who want measurable AI outcomes — not consulting decks. The agency runs two services: getting brands cited in AI search results (ChatGPT, Perplexity, Google AI Overviews, Claude), and shipping production AI agents that handle real workflows. Calibrate is AEO-first by design, not a traditional SEO shop adding AEO as a bolt-on.

An AEO audit is the first thing Calibrate runs for any brand, and it is the step most agencies skip. The reason it comes first is simple: you cannot improve a citation rate you have never read. Before a single page is written, you need to know whether AI engines name your brand, on which questions, in what position, and against which competitors. Everything after that is guesswork without it.

This guide is the working method, not a theory piece. It walks through what an audit measures, how to build the prompt set that tests your category, how to read citation rate and answer position, how to find the third-party sources the engines actually trust, and how to map the competitors who own the answers you want. It covers the tools, the time, the cost, and what to do with the results.

By the end you will be able to run a basic AEO audit yourself, or judge whether the one an agency is selling you is real. The test of a real audit is whether it produces a one-page baseline you can act on: where you are cited, where you are absent, and who is cited instead. If it produces a slide deck instead of a baseline, it is theatre.

Written by Prashant Kochhar · Calibrate · Updated June 2026

Table of Contents

1. What is an AEO audit?

2. Why run an AEO audit before any content?

3. What does an AEO audit measure?

4. How do you build the prompt set to test?

5. How do you read citation rate and answer position?

6. How do you find which sources the engines trust?

7. How do you map competitors in AI answers?

8. What tools do you need to run an AEO audit?

9. How long does an AEO audit take and what does it cost?

10. What do you do with the audit results?

11. Related Guides from Calibrate

Last updated: June 2026 · Next update: October 2026

What is an AEO audit?

An AEO audit is a structured measurement of how often and how prominently AI engines cite your brand across the questions buyers ask in your category. It reads five engines — ChatGPT, Perplexity, Claude, Google AI Overviews, and Microsoft Copilot — and produces a baseline: your citation rate, your position inside answers, the sources the engines trust, and the competitors named instead of you.

It is a diagnostic, not a content exercise. Nothing gets written or changed during an audit. Its only job is to replace assumptions with numbers, so the work that follows aims at measured gaps rather than guesses. This is the Audit phase of the wider method described in the Citation Architecture method, and it is the foundation every later phase depends on.

The output should fit on a page. A real audit ends with a baseline a founder can read in two minutes and act on, not a forty-slide document that restates the obvious. If you remember one thing, it is that an audit is judged by the clarity of the baseline it produces.

Why run an AEO audit before any content?

You run an audit first because content built without a baseline aims at nothing measurable, which is the most common way AEO budgets are wasted. The instinct is to start producing, because producing feels like progress. Without a baseline, that energy tends to reinforce queries you already win and ignore the ones a competitor owns.

The urgency comes from where buyers now go. According to Gartner's forecast on search behaviour, a quarter of traditional search volume is moving to AI assistants by 2026, so a brand that has never measured its position in AI answers is unmeasured on a fast-growing slice of its market. An audit closes that blind spot before any money is spent on content.

The shift is not confined to one market either. According to a16z's ranking of the most-used consumer AI apps, India and the UAE sit among the top markets for every major assistant, so for a brand in those regions the questions buyers ask AI are a present concern rather than a future one, and worth reading today rather than next year.

The audit also changes the quality of every brief that follows, because each one can point at a specific gap rather than a guess. That is why a content-first agency that skips measurement tends to produce volume without movement, and why the cheapest, highest-return move in AEO is to measure before you build.

What does an AEO audit measure?

An AEO audit measures four things: your citation rate, your position within answers, the trusted sources feeding each engine, and the competitors cited against you. Each is read per engine, because a position on one is no guide to another, and a single-engine view hides where you are actually losing.

Citation rate is the headline: the share of relevant answers that name your brand. Position tells you whether you are listed late or recommended first. The source map shows which third-party pages the engines pull from for your category. Competitor citations show who owns the answers you want. Together they turn a vague sense of presence into a precise picture.

The reason all five engines must be read is that they retrieve differently and reward different signals, a point detailed in the AI engines that decide your visibility. A brand can look healthy on Perplexity and be absent on Google AI Overviews, and only a multi-engine read catches that. How these numbers are calculated and tracked over time is covered in how to measure AEO.

How do you build the prompt set to test?

You build the prompt set by listing the real questions buyers ask in your category, in their words, weighted toward commercial intent. The prompt set is the heart of the audit, because the engines are tested by asking them the questions your buyers actually ask, then recording whether your brand appears.

Start from intent, not keywords. A keyword list gives you fragments; a prompt set gives you the full questions a buyer types into an assistant. Cover the main buying questions, the comparison questions, the trust questions, and the category-defining questions. Aim for breadth across the buyer journey, then weight the analysis toward the high-intent prompts closest to a purchase.

Building the set from questions rather than keywords is the same reason keyword lists fail for AI search: the engines answer questions, so you test them with questions. A useful set runs to a few dozen prompts for a focused brand, enough to cover the journey without diluting the read with queries no buyer asks.

How do you read citation rate and answer position?

You read citation rate as the share of your prompt set where the brand is named, and position as where in each answer the mention falls. A brand named in fifteen of fifty prompts has a thirty percent citation rate on that set. Reading it well means looking past the single number to where the mentions cluster and where the gaps are.

Position matters because being named is not the same as being recommended. A brand listed last, after three competitors, has a weaker citation than one named first as the top pick. Record both, per engine, so you can see not only whether you appear but how strongly. The pattern of strong topics and missing topics is what turns the read into a plan.

The honest read shows both halves. In Calibrate's own work on a retail brand, the audit found strong citations on premium apparel and complete absence on adjacent categories, and showing both is what made the plan credible. The proof of that read, with real numbers, is in the Cobbled Climbs case study.

How do you find which sources the engines trust?

You find the trusted sources by recording, for each answer, which pages the engine cited or drew from, then looking for the sites that recur across your category. Answer engines lean on a narrow pool of sources they treat as reliable, and that pool is often more important than your own pages, because being referenced there is what gets you named.

The pattern is consistent across categories. For most commercial questions, the recurring sources are a Wikipedia entry, one or two high-signal community threads such as Reddit, a respected comparison article, and a handful of category authorities. Recording which of these the engines pull from tells you where the real work sits, on your pages or on third-party sources.

Google's own systems work this way, grounding answers in retrieved sources. According to Google Search Central's guidance on AI features, its generative features retrieve indexed content and favour original, first-hand material, which is why earning a place in the trusted pool matters as much as polishing your own site. The method for earning community citations without being removed is covered in how to earn a Reddit citation.

How do you map competitors in AI answers?

You map competitors by recording which brands are named in the answers where you are absent or ranked below them, then counting how often each appears across your prompt set. This turns a vague sense of who the competition is into a ranked list of who actually owns the AI answers in your category, which is frequently different from who you consider your market rivals.

The useful output is a share-of-voice picture: for each priority query, who is named, in what order, and through which source. A competitor that owns a query through a single strong Reddit thread is a different problem than one that owns it through ten authority articles, and the map tells you which.

The queries where no brand clearly leads are the quickest wins, because the answer is still contestable. Mapping competitors this way feeds directly into prioritisation: you fight first where a citation is both valuable and winnable, the decision logic of the Architect phase in the Citation Architecture method.

What tools do you need to run an AEO audit?

You need a way to query each engine systematically and a way to record and track the results, which ranges from manual testing to a dedicated AEO tracking platform. A basic audit can be run by hand: ask each engine every prompt in your set, record whether and where your brand appears, and tabulate the results in a spreadsheet. It is slow but real.

At scale, manual testing breaks down, because answers vary between runs and across five engines, and a one-off read does not show trends. Calibrate uses Searchable to track citation rate, position, and share of voice across engines continuously, which turns a one-time snapshot into a monitored baseline. The tool you choose matters less than running every prompt across every engine and recording it consistently.

For a one-time read on a small brand, manual testing and a spreadsheet are enough to produce a usable baseline. For an ongoing programme where you need to see the trend month over month, a tracking platform is the practical choice, since the value of an audit compounds when it becomes a monitored number rather than a single measurement.

How long does an AEO audit take and what does it cost?

A focused AEO audit takes one to two weeks and, run as a fixed-scope engagement, costs in the region of 1,500 to 3,000 US dollars, depending on category breadth and the number of engines. A manual self-run audit costs only time, typically a few days of disciplined testing and recording for a small prompt set.

The cost scales with the size of the prompt set, the competitiveness of the category, and whether you are reading a snapshot or setting up continuous tracking. The figures below are indicative and founder-led, with the work running directly rather than through an account layer.

The return on an audit shows up on every later decision, because it stops you from spending a content budget blind. A single accurate baseline can redirect months of content work away from queries you already win and toward the ones a competitor currently owns. Treated correctly, the audit is the cheapest part of an AEO programme and the part that makes the rest efficient. Whether your category justifies a programme at all is the question in does my business need AEO.

What do you do with the audit results?

You turn the audit into a ranked plan: defend the queries you own, prioritise the winnable gaps, and decide which missing queries are worth entering. The baseline is not the deliverable; the decisions it enables are. A good audit ends with a short list of where to act, in order, with the move chosen for each.

The translation is direct. Strong queries get defended and refreshed. Queries where a competitor leads and the answer is contestable get prioritised for the build. Missing queries get a yes or no decision on whether they are worth the work. That ranked list is the input to the Architect phase, where strategy is set.

From here the work moves into building, and the audit becomes the baseline every result is measured against. If you want a professional read across all five engines before deciding anything, Calibrate runs it as a fixed-scope AEO audit, and the full picture of how the audit feeds the rest of a programme is on the services page. The discipline that makes it pay off is using the baseline, not filing it.

Frequently Asked Questions

What exactly does an AEO audit check?

An AEO audit checks how often and how prominently AI engines name your brand across the real questions buyers ask in your category. It reads five engines, ChatGPT, Perplexity, Claude, Google AI Overviews, and Microsoft Copilot, and records your citation rate, your position inside each answer, the third-party sources the engines trust, and the competitors named instead of you. The output is a one-page baseline showing where you are cited, where you are absent, and who is cited in your place, so later content aims at measured gaps rather than guesses.

Can I run an AEO audit myself?

Yes, a basic one. List the real questions buyers ask in your category, ask each of the five engines every question, and record whether and where your brand appears in a spreadsheet. That produces a usable snapshot for a small brand. The limits are that answers vary between runs and a one-off read shows no trend, so manual testing does not scale well across five engines and dozens of prompts. For an ongoing programme, a tracking platform that monitors citation rate continuously is the practical step up from a manual audit.

How many prompts should an AEO audit include?

For a focused brand, a few dozen prompts is usually enough to cover the buyer journey without diluting the read. The set should span buying questions, comparison questions, trust questions, category questions, and any region-specific queries, then weight the analysis toward the high-intent prompts closest to a purchase. More prompts add resolution but also noise, so the goal is coverage of the questions buyers actually ask rather than an exhaustive list of every phrasing. Quality and intent matter more than raw count.

How is an AEO audit different from an SEO audit?

An SEO audit checks technical health, rankings, and on-page factors that affect your position in a list of links. An AEO audit checks whether AI engines cite your brand inside their answers, measuring citation rate, answer position, trusted sources, and competitors across five engines. The two overlap on technical foundations but diverge on the target: one measures rankings, the other measures citations. Running an SEO audit tells you nothing about your AI visibility, which is why an AEO audit is a separate, citation-focused diagnostic rather than an add-on to an SEO report.

How often should I run an AEO audit?

Run a full baseline audit once to establish where you stand, then re-read the key metrics monthly as part of an ongoing programme. AI answers shift as engines retrain and competitors act, so a single audit is a snapshot that ages. Monthly tracking turns it into a trend you can manage, showing which citations held, which slipped, and where new gaps opened. For a brand not yet running a programme, an annual or quarterly re-audit is the minimum to avoid drifting blind, but monthly is where the method compounds.

What does an AEO audit cost?

A self-run manual audit costs only your time, typically two to four days for a small prompt set. A fixed-scope professional audit generally runs in the region of 1,500 to 3,000 US dollars, depending on category breadth and the number of engines and prompts covered. Ongoing tracking is usually folded into a wider programme rather than priced separately. The cost is modest relative to the content budget it protects, since the audit is what stops that budget from being spent on queries you already win or cannot realistically take.

Do I need a tool, or is manual testing enough?

For a one-time read on a small brand, manual testing and a spreadsheet are enough to produce a usable baseline. A tool becomes necessary when you need trends rather than snapshots, because answers vary between runs and tracking five engines by hand each month is impractical. Calibrate uses a dedicated platform to monitor citation rate, position, and share of voice continuously, which turns the audit from a single measurement into a managed number. Start manual if you are testing the idea, and add a tool when you commit to a programme.

What should the audit deliver at the end?

A real AEO audit delivers a one-page baseline and a ranked action list, not a long slide deck. The baseline shows your citation rate and position per engine, the trusted sources in your category, and the competitors cited against you. The action list translates that into decisions: which queries to defend, which winnable gaps to prioritise, and which missing queries to enter or skip. If an audit produces a thick document that restates the obvious without a clear baseline and next steps, it is presentation rather than diagnosis.

Related Guides from Calibrate

• What Is AEO? Answer Engine Optimization Explained — the discipline the audit serves.

• The Citation Architecture Method — where the audit sits as phase one.

• How to Measure AEO — the metrics the audit reads, tracked over time.

• The AI Engines That Decide Your Visibility — why all five engines are read separately.

• How Cobbled Climbs Got Cited for Premium Cycling in India — an audit baseline turned into results.

• Schema for AI Engines vs Schema for Google — the structured-data work the audit points to.