How to choose an AEO agency: 8 things to check

BrandurePublished Last updated

To choose an AEO agency, check how it prices and measures the work before you look at what it promises: per-engine reporting, a logged-out baseline, and a plan weighted towards earned third-party placements are the clearest signs it knows what moves AI answers. Walk away from anyone guaranteeing a position or a citation in ChatGPT, Gemini or Google AI Overviews, because no one controls those.

Brandure is an AEO agency, so we have an interest here. These are the same checks we’d want a prospective client to run on us.

The eight checks below are ordered by how much they protect you, starting with the money.

1. How does the agency price the work, and what are you actually paying for?

AEO is sold in a handful of models: a monthly retainer, a one-off audit, a fixed sprint, a software subscription, or a hybrid of audit-then-retainer or tool-plus-operator. HubSpot’s September 2026 roundup lists agency retainers from around $249 a month at entry level to $10,000–$15,000+ for full-service programmes, diagnostic audits from free to a few hundred dollars with one-off audit sprints from around $3,000, and self-serve tracking tools from about $29 to $489 a month (HubSpot). HubSpot sells its own AEO tool, so read those as market examples rather than benchmarks.

The useful question is what sits inside the fee. Ask for the split between measurement, on-site work and off-site work, and whether link building, original research or PR are extra. HubSpot notes these are often add-ons.

If most of the fee is a tracking dashboard you could subscribe to yourself, you’re paying a reseller margin, not buying a service.

2. How long is the contract, and what happens when you leave?

HubSpot notes some agencies ask for a three-month minimum. That’s reasonable for work that needs time to show. What matters more is the exit:

  • Notice period. How long after you decide to stop?
  • Ownership. Do you keep the content, the profile logins, the review platform accounts and the tracking history?
  • What decays. AI answers move month to month. In our September 2026 testing, ChatGPT’s list of best Dubai law firms changed substantially from August: the previous first-place firm dropped out entirely, and two firms moved up from fourth and eighth into the top two. A position has to be held, so ask what stops working when the retainer does.

An agency that can’t answer the third question hasn’t thought about it.

3. Does the agency report each AI engine separately?

It should. ChatGPT, Perplexity, Gemini, Google AI Overviews, Claude and Copilot draw on different sources and rarely name the same businesses. In our Dubai testing across seven categories, overlap between engines came back low in most categories, with real agreement only where a strong third-party authority (such as a recognised directory or ranking) existed. A separate practitioner study found Claude’s citations matched ChatGPT’s only 8% of the time on identical prompts (Search Engine Land).

A single blended “AI visibility score” averages those positions together and tells you nothing you can act on. You could be first on one engine and missing from another.

Ask also how many times each prompt is run. We asked ChatGPT the same aesthetic clinics question twice, minutes apart: first place held, and half the list below it changed.

Position Run 1 Run 2
1 Clinic A Clinic A
2 Clinic B Clinic G
3 Clinic C Clinic B
4 Clinic D Clinic D
5 Clinic E Clinic H

Single-run reports are snapshots, not positions.

4. Will they take a baseline before doing any work?

Without a baseline there’s no way to prove anything moved. A good baseline is:

  • Logged out, on a clean browser profile. Signed-in results are personalised.
  • Location-set to your market.
  • Repeated, not single-run.
  • Source-recorded: which publications or directories each engine cited, not just whether you were named.

Google’s own advice for hiring search help includes asking “what results should I expect, over what timeframe, and how do you measure success?” (Google Search Central). The baseline is where that answer starts. Our guide to checking whether ChatGPT recommends your business sets out the method step by step, so you can run your own before talking to anyone.

5. Is the plan weighted towards earned placements or your own website?

This is the check that separates AEO from rebadged content marketing.

Engines increasingly ground answers on sources that look independent. In our testing, one Dubai SaaS company held first place on ChatGPT in both August and September, but by September ChatGPT was backing that position with three third-party publications instead of the company’s own “best of” article. ChatGPT had stopped citing self-published lists in every category we tested. Perplexity still cites them.

Where a category has an authority layer, it dominates: Legal 500 was ChatGPT’s only source for Dubai law firms. Where trade press exists, ChatGPT spreads across it. Where neither exists, review volume decides.

So ask which third-party sources the engines cite in your category today, and what share of the plan goes on getting you into them. If the answer is mostly blog posts on your own site, the plan is aimed at the wrong layer for most engines.

6. Do they do entity and profile work?

Engines have to know who you are before they can name you. Two examples from our data:

  • Entity confusion. One Dubai clinic appears under two spellings of its name, with two Google listings. Perplexity treated it as one business; ChatGPT and Gemini each picked one spelling and ignored the other. Inconsistent names and duplicate listings split your visibility in two — an identity problem, fixed with consistent names, descriptions and profile data, not with more content.
  • Google’s own feeds. Gemini’s answers mapped onto Google products: the Google Hotels module for hotels, Google Maps for local services, Shopping cards for manufacturers, YouTube for universities. For a local business, Google Business Profile work is AEO work.

An agency that never mentions profiles, directories or business identity is skipping the cheapest wins.

7. Are their timelines realistic and tied to specific engines?

Be wary of any fixed timeline quoted before a baseline exists. Speed differs by engine and lever. Profile and identity fixes are within your control and quick to ship; earned coverage and directory standing take longer to build. HubSpot’s guidance is to “expect early signals, not certainty, in the first 60 to 90 days” (HubSpot).

A credible plan sequences the faster levers first so you can see motion, and says which engines it expects to move first and why.

8. What are the red flags?

  • Guaranteed positions or citations. Google’s line on SEO applies here too: “No one can guarantee a #1 ranking on Google” (Google Search Central). For AI Overviews it’s especially hollow. Across three identical searches we saw Google name the same three manufacturers each time but rotate which source it cited.
  • llms.txt or special schema sold as the core deliverable. Google’s guidance says you don’t need special files, markup or schema for its AI features, and that Google Search ignores llms.txt (Search Engine Journal, May 2026; Search Engine Journal, June 2026). It’s a half-day housekeeping task, not a retainer.
  • One blended score, one engine, or one run. See check 3.
  • No sources in the report. Without knowing who the engine cited, you can’t tell which lever to pull.
  • Charging separately for “AEO” and “GEO”. They’re the same work.
  • Case studies they can’t show. Ask for the prompt set, the dates and the logged-out conditions behind any claimed result.

The checklist at a glance

Check Ask Good answer Red flag
Pricing What’s inside the fee? Clear split of measurement, on-site and off-site work Mostly a tool subscription
Contract and exit What do I keep if I leave? You own content, accounts and data Vague notice, agency holds logins
Reporting How do you report engines? Per engine, repeated runs, sources listed One blended score
Baseline When do you measure the start? Before any work, logged out, location set No baseline, or signed-in results
Earned vs own-site Which third-party sources will you target? Names the sources your category’s engines cite Only blog posts
Entity work How do you handle profiles and identity? Profile, directory and naming audit Not mentioned
Timelines Which engines move first, and why? Sequenced by engine and lever Fixed promise before a baseline
Guarantees Can you guarantee a position? No, and explains why Yes

FAQs

How much should an AEO agency cost? It depends on scope and market. HubSpot’s 2026 roundup shows agency retainers ranging from a few hundred dollars to over $15,000 a month. Compare what’s inside the fee, not the headline number.

Should I hire an AEO agency or use a tracking tool myself? A tool tells you where you stand. It doesn’t get you into the directories, press and profiles that change the answer. If you have someone in-house to do that work, a tool may be enough.

Is a three-month minimum contract reasonable? Usually, yes. Answers take time to move and some engines update slowly. What matters more is a clear exit and that you own what’s been built.

Can an AEO agency guarantee I’ll appear in ChatGPT? No. No agency controls what an AI engine names. A credible agency commits to process, measurement and the sources it will target, not to a position.

Do I need separate agencies for SEO and AEO? Not necessarily. Most engines search the web before answering commercial questions, so the two overlap. What you need is someone who measures and reports each AI engine separately. Our guide to answer engine optimisation explains how the engines differ.


Sources

All guides →