AEO is a sequence, not a checklist. The order is the whole point.

Your customer had a burst pipe. They asked ChatGPT who to call. It named three plumbers in one sentence. You were not one of them, and you will not know unless you check.

This is the quiet part of “AI search.” A year ago, ranking well on Google was a fair sign you were being found. Now an AI answer often sits on top of the results and names two or three businesses directly. If you are not one of the names, your ranking underneath does not save you.

Most of the advice sold to fix this is a pile of tactics with no order to it. That is the mistake. The work is not new. The sequence is what is missing.

What this feels like from your side

We hear the same things from roofers, plumbers, HVAC companies, dentists and law firms, almost word for word:

• “We rank fine on Google, but the AI names someone else.” A ranking screenshot tells you nothing about whether an AI pulled your page into its answer.
• “I have no idea what the AI actually says about my business.” Most owners have never typed their own service into ChatGPT or Perplexity and read the result.
• “I bought ‘AI SEO’ and got a list of tasks with no plan.” Tactics with no order and no starting measurement cannot be shown to have worked.
• “Leads show up already decided.” They asked an AI first, picked someone, and only then started calling.
• “A competitor with worse work keeps getting named.” Usually because their business facts are clean and easy for a machine to confirm, and yours are not.
• “I can’t tell my team if any of it is working.” There is no number to point at.

Why most “AI SEO” fails: it is a menu, not a recipe

A menu lets you pick the parts you like. A recipe has steps that have to happen in order, or the dish does not come out. AEO, short for Answer Engine Optimization, is the work of getting named in AI answers. It is a recipe.

Here is why the steps depend on each other:

• Google says the basics still decide it. Google’s own guide for its AI features says the established foundations of SEO “continue to be relevant,” and that its AI answers are built on the same Search index and quality systems as normal results.¹ There is no separate AI ranking system to buy your way into.
• ChatGPT and Perplexity read the live web and cite named sources. ChatGPT’s search answers with current information from the web and links to the pages it used.² Perplexity builds every answer from live sources and shows them inline.³ Same web. Same pages. If your page cannot be found and trusted, none of these three can name you.
• A machine will not state a fact it cannot confirm twice. If your business name or address reads one way on your site and another way on a directory, an AI hedges, or names the competitor whose facts line up.

So the work is ordinary. Do it in the wrong order and you waste most of it.

The five steps, in order

Step 1: Baseline (week 1–2)

Write down 20 to 30 real questions your customers ask before they hire someone. Not keywords. Questions. “Who does emergency roof repair in Surrey?” “Best HVAC company near me for a heat pump.” “How much does a kitchen remodel cost in Langley?”

Type each one into Google’s AI Mode, ChatGPT and Perplexity. For each, write down two things: did it name you, and did it name a competitor. That list is your “before” picture. It is the only thing that lets step 5 mean anything, and it exists whether the results flatter you or not. It is the same rule we hold ourselves to with a baseline map scan before any local SEO work starts.

Step 2: Entity (week 2–4)

“Entity” just means the record of who your business is. Same name, same address, same phone number, same words for your services, everywhere: your website, your Google Business Profile, and every directory an AI might read.

This is the step people skip and it is the one that moves the needle most. An Ahrefs analysis found that plain brand mentions across the web lined up with AI visibility far more strongly than backlinks did.⁴ In our own baseline audits, most businesses have their services named three different ways across three sites without realising it. Fix that first and the later steps have something solid to stand on. More on the mechanics: local business entity SEO.

Step 3: Content (week 4–6)

Answer first. Put the plain answer to the question in the first one or two sentences of the page, then explain underneath. Do not write 800 words before you answer it.

An AI reading your page is trying to lift a short, clear, self-contained answer. If it has to dig for it, it grounds on a page where the answer sits right at the top instead. This is not a special “AI writing style.” Google says its systems understand normal human language.¹ It is just good structure. Every page on this site opens with a “Quick answer” block for exactly this reason.

Step 4: Technical (week 1–3, runs alongside)

The page has to load fast, be crawlable, and sit in Google’s regular index. Google’s AI answers are “grounded.” Before writing, the system pulls real pages from that same index, the way an inspector walks the house before writing the report.¹ If Google cannot read the page, the AI has nothing to ground on and cannot cite you. No AI trick fixes a page a crawler cannot reach.

Step 5: Measure (every month)

Re-run the exact same questions from step 1. Count how many now name you. Watch that number across months, not weeks.

Two honest points. First, you are counting citations (mentions in the answer), not ranking positions. Second, it is a roll of the dice each time: the same question can give a slightly different answer twice in a row, which is why the trend over three months matters and a single check does not.

The mistakes that don’t show up until you check

The detector covers the ones people argue about. These are the quiet ones, and nothing on your site looks broken:

• Chasing mentions on any site that will have you. AI answers lean on the directories and review sites customers actually use, not a scattershot of low-quality links.
• Chopping pages into tiny “AI chunks.” Google says its systems already understand a full page.¹ Disconnected fragments just read worse for the human who lands on them.
• Measuring rankings instead of citations. A rank tracker cannot see inside an AI answer. It never could.
• Judging it at week two. Entity and content work compounds over a couple of months. Checked too early, every campaign looks stalled.
• Never re-running the questions. If you do not repeat step 1, you have no idea whether step 5 moved.

Your crew is already making the proof

Here is the part that costs established businesses the most. Your crews finish real jobs in real neighbourhoods every week. That is exactly the specific, checkable, first-hand evidence an AI wants to ground on, and almost none of it ever reaches a page. It lives on a phone, in a text thread, in your memory of a good week. Google knows you work there; your website doesn’t, and neither does the AI.

Getting that proof onto the site automatically, in a dated and mapped form a machine can read, is the whole idea behind Map Check-Ins: field photos, timestamped and pinned to a location, published to your website and Google Business Profile as jobs happen. It is not an AI trick. It is step 2 and step 3, done for you every week, using work your crew already did. $299 a month, alongside the roadmap, not instead of it.

What you actually want out of this

Strip away the noise and the goal is short:

• When someone asks an AI for the best in your trade and city, your name is in the answer.
• A list of your customers’ real questions with a yes-or-no next to each: are we named?
• Knowing which of your pages get cited, so you can make more like them.
• AI-referred visitors who arrive already half-sold, because the answer vouched for you.
• One set of fundamentals that works for Google, ChatGPT and Perplexity at once, not three projects and three invoices.
• Your finished jobs turning into proof on their own.
• A single number you can watch go up each month.

The honest part

None of this is a guaranteed position, and anyone who promises one is not worth hiring. Google’s own guidance says so.¹ AI answers are probabilistic: some months the number will not move, and we will say that at the review instead of waiting to be asked. We take a baseline before you pay for anything, and if your market cannot be moved for what you can spend, the honest recommendation is to work the free Map Pack Check yourself.

One more thing: we take one client per industry, per area. Two roofers in the same city cannot both be moved to the top of the same answer, and an agency that takes both is charging two people to compete with each other. When a territory is taken, we say so and decline.