AEO Services: The Eight Things You Actually Receive

Updated September 2026 · Written and maintained by the Progression Agency strategy team

AEO services from Progression Agency are answer engine optimization sold as artefacts rather than activities — the reports, rewrites and baselines you keep, and how to verify every one of them yourself.

On this page · 14 sections

What AEO services actually deliver
What is included, and what is deliberately not
How the work is sequenced across six months
The two tests that happen before we quote
How each deliverable is evidenced
What the structural work actually changes on a page
Which services you actually need
Is this market open to a smaller supplier?
What it costs
Everything else we have written on search, AI and getting found
The deliverables, with the evidence that proves each one
In scope, out of scope, and the one genuinely blurred line
How AEO services differ from the SEO services you may already buy
Service levels, and who each one suits

The short answerAEO services deliver eight artefacts in six weeks: a crawler access report from your server logs, a robots audit, a render report per template, a dated prompt baseline, passage rewrites, an entity sheet, a schema diff, and monthly measurement against that baseline. Every one is independently checkable. Not included: article volume, keyword research, rank tracking or any guarantee of citation. Two tests before quoting decide the scope — scriptless rendering, and your robots.txt against the AI crawler list.

Search volumes, CPC and domain authority figures measured September 2026. Published pricing bands are attributed on our AEO pricing page.

What you receive, week by week — Eight artefacts. Every one is a file you keep, and every one is checkable by someone else.

What AEO services actually deliver

Eight artefacts in the first six weeks, and every one of them is a file you keep and can check yourself: a crawler access report from your server logs, a robots audit, a render report per template, a dated prompt baseline, passage rewrites, an entity sheet, a schema diff, and then a monthly measurement against the baseline.

That list is deliberately concrete because this discipline has an evidence problem. A great deal of what is sold as answer engine optimization arrives as a dashboard and a narrative. Neither can be verified by anyone other than the supplier who produced it.

Note what is not on the list. No content calendar, no keyword map, no rank report. Those belong to a different channel, and two of them actively make AEO performance worse.

Access report — Artefact. AI crawler hits from your logs, by bot, with status codes.

Render report — Artefact. Scripted vs scriptless word count, per template.

Prompt baseline — Artefact. Fixed wording, verbatim answers, dated.

Passage rewrites — Artefact. Before and after, section by section.

Entity sheet — Artefact. One description, plus every source contradicting it.

Schema diff — Artefact. What the markup claims versus what the page says.

The crawler access report

A table from your own logs showing which AI bots requested which URLs, what status code they got, and how often. It is the only way to know whether access is real rather than theoretical, and you can re-run it without us.

The render report

Every commercial template fetched twice — once as a browser, once with scripts disabled — with the word counts side by side. This single document usually explains why a site is absent from AI answers.

The prompt baseline

A fixed set of buyer questions, run once, with the verbatim answers stored and dated. Without this, nothing claimed later can be proven, because answers vary between runs and memory is not evidence.

The passage rewrites

Before and after, section by section. The change is almost always the same: move the answer to the first sentence of the section that promises it.

The entity sheet

One canonical description of the business, plus a list of every source currently contradicting it. Tedious work that stops assistants saying wrong things about you.

The schema diff

What your structured data claims, set against what the visible page actually says. Markup that contradicts the copy is worse than no markup at all.

What is included, and what is deliberately not

In scope: crawler access, rendering diagnosis, passage restructuring, schema reconciliation, entity consistency, corroboration and prompt-set measurement. Out of scope: article volume, keyword research, rank tracking and any form of guarantee.

The honest boundary is rendering implementation. We specify the fix and test it, but actually changing how your application renders is usually an engineering job that belongs with your developers. Pretending otherwise is how AEO quotes become wrong by an order of magnitude.

What is in scope and what is not — The ‘part’ row is the honest one: we specify the rendering fix and test it, but implementation usually belongs with your engineers, and that changes the price substantially.

Article volume — Not included. Publishing more is not the lever here.

Keyword research — Not included. No ranked list means no per-keyword target.

Rank tracking — Not included. Cannot see inside an answer.

Guarantees — Not included. Nobody controls model output.

Paid placement — Not included. No inventory exists to buy.

Engineering delivery — Not included. We specify and test; your devs usually implement.

How the work is sequenced across six months

Front-loaded and cheapest-first. Month one is access and baseline, month two is rendering, months three to five are structure, entity and corroboration, and month six is the first point at which the measurement says anything reliable.

The shape of the hours inverts over that period. Month one is overwhelmingly technical; by month six most of the effort is corroboration, which is the slowest signal and the one with the longest half-life.

A six-month engagement, month by month — Month six is the first point at which citation frequency says anything reliable. Any supplier claiming a verdict in month two is reading variance.

Where the hours go in an AEO service month — The shape inverts over six months. Front-loaded technical work gives way to corroboration, which is the slow half and the durable one.

The two tests that happen before we quote

We fetch your commercial templates with JavaScript disabled, and we read your robots.txt against the current AI crawler list. Both take an afternoon, and together they decide the scope and therefore the price.

This is not a courtesy. Rendering health is the single largest cost variable in an AEO engagement, and a quote produced without testing it is a guess with a number attached.

The two tests we run before quoting — We do not quote before these. A price produced without them is a guess dressed as a proposal.

Robots.txt — Technical. Allow the five crawlers that matter.

Rendering — Technical. Content must survive without JavaScript.

Status codes — Technical. 403s and soft 404s to AI bots are common and invisible.

Redirects — Technical. Chains lose retrieval more often than they lose rankings.

CDN and WAF — Technical. Bot protection blocks AI crawlers by default more often than you think.

Log access — Technical. Without it, crawler claims are unprovable.

How each deliverable is evidenced

Every artefact is independently checkable. Log extracts you can re-run, word counts you can reproduce with a single command, verbatim stored answers, and links to the actual third-party pages rather than a count of them.

We are explicit about this because the failure mode in AEO reporting is not dishonesty so much as irreproducibility. Answers vary between runs, so a screenshot proves nothing and a citation count with no stored answers cannot be audited by anyone, including us.

How each deliverable is evidenced — Every green row is independently checkable. That is deliberate: this discipline has a fabrication problem, and stored verbatim answers are the cure.

Citation frequency — Measured. How often you are named across the prompt set.

Share of voice — Measured. You versus competitors in the same answers.

Sentiment — Measured. How you are described, not just whether.

Prompt coverage — Measured. Which buyer questions you appear on at all.

Crawler hits — Measured. Whether access is real, from logs.

Render coverage — Measured. Templates a machine can actually read.

Why verbatim answers are stored

Non-determinism is real. The same prompt can return different sources an hour apart. Storing the answer text is the only way to distinguish a genuine change from variance.

Why the prompt wording is fixed

Changing a single word changes which sources are retrieved. A prompt set that drifts produces a trend line that means nothing.

Why rankings are reported separately

Merging them is how weak AEO work hides behind ordinary search gains. If the two are on the same chart, one of them is decoration.

Why we report sentiment, not just presence

Being described wrongly is worse than being absent. A brand named alongside an inaccurate claim has a problem a citation count will never surface.

What the structural work actually changes on a page

One thing, mostly: the answer to a heading moves into the first sentence of the section that promises it. Retrieval selects passages, so an answer assembled across three paragraphs and a bullet list has nothing clean to lift.

The secondary changes are comparisons expressed as tables, definitions stated plainly rather than approached gradually, and numbers carrying their sources inline. All three survive summarisation better than prose does.

Heading answers itself — Structure. First sentence of the section resolves the heading.

One idea per section — Structure. Retrieval lifts a passage, not a page.

Tables for comparisons — Structure. Extract cleanly and survive summarisation.

Definitions stated plainly — Structure. Definition prompts are the highest-citation query type.

Numbers with sources — Structure. Unsourced statistics are the first thing dropped.

The commonest structural failure

The heading asks a question and the section answers it in paragraph four, after context. Human readers tolerate this. Retrieval does not.

Why tables extract well

A table row is already a self-contained fact with its dimensions labelled. It is close to the shape retrieval wants.

Why definitions matter disproportionately

Definition prompts are among the highest-citation query types, because a model answering ‘what is X’ needs a clean sentence to quote.

Why unsourced numbers get dropped

A statistic without attribution is the first thing a cautious model declines to repeat. Sourcing inline is cheap insurance.

Which services you actually need

Not every site needs the full programme. If your category is question-led and your site blocks machines, you need all of it. If demand is brand-led and the site is readable, you need a baseline and very little else this year.

We give the second answer regularly. It is a poor sales tactic and an accurate one.

Which services you actually need — Bottom-left is a real answer and we give it. If demand is brand-led and the site is readable, there is nothing here worth buying this year.

Best fit — B2B SaaS. Long research cycles, assistant-using buyers.

Best fit — Professional services. Question-led, high-value discovery.

Best fit — Healthcare. Question-shaped demand; wrong answers carry real risk.

Best fit — Fintech. Comparison-heavy, corroboration-sensitive.

Best fit — Manufacturing. Specification queries suit passage retrieval.

Poor fit — Local trades. Listings and reviews still return more today.

Full retainer

Question-led category, real technical debt. Everything in the deliverables list.

Audit plus fixes

Question-led category, clean site. Buy the diagnosis and the structural work, skip the retainer until there is something to measure monthly.

Technical sprint only

Brand-led demand but machines are blocked. Fix access and rendering, then stop. It is cheap and it removes a genuine risk.

Baseline and monitor

Clean site, brand-led demand. Record a prompt set, check it monthly, and spend the budget somewhere it returns more this year.

Is this market open to a smaller supplier?

Right now, yes, unusually so. A DA-7 domain holds a position on the second page for ‘aeo services’, a term with a $38 cost per click, and the median page-one domain authority is around thirty.

That is a window rather than a permanent condition. Established brands are still treating AEO as a blog subject rather than a service line, which is why specialists with modest authority are currently outranking them.

Domain authority currently ranking for 'aeo services' — A DA-7 domain holds a position on page two of a commercial term with a $38 cost per click. This family is genuinely open right now.

Cost per click on the AEO services family — Cost per click is the market pricing the intent. At $25-$70 a click these are buyers, not researchers.

What it costs

Published US retainers for this work run roughly $1,500 to $20,000 a month and a one-off audit $1,000 to $4,000. Your number depends on template count and rendering health far more than on the market.

Buy the audit first unless you already know your rendering is clean. It converts assumptions into a scope, and it occasionally concludes that you do not need the rest.

What an AEO service is actually optimising toward — Two tests decide the price, eight artefacts prove the work, and nothing can be bought. That is the whole commercial shape of this service.

Start with the two tests

We will fetch your commercial templates with JavaScript disabled and check your robots.txt against the current AI crawler list, then tell you what we found — including if the honest answer is that you do not need a retainer this year.

/ai-visibility-audit

Want this done for your site?We deliver the eight artefacts on this page in six weeks, each one independently checkable by you.

Get a free proposal

Everything else we have written on search, AI and getting found

Getting found in search

AI, AEO and what is changing

Paid media and lead generation

Websites and design

Choosing and working with an agency

By industry and by situation

Not sure which of these applies to you?Tell us the situation and we will say plainly what we would do first, and what we would not.

Talk it through

The deliverables, with the evidence that proves each one

Eight deliverables, each paired with the artefact that proves it was actually done.

Every artefact and how you check it yourself
Deliverable	When	What you receive	How you verify it independently
Crawler access report	Week 1	Log table by bot, URL and status code	Re-run the same grep on your own logs
Robots audit	Week 1	Scored robots.txt plus exact lines to change	Read the file; the change is visible
Render report	Week 2	Scripted vs scriptless word count per template	curl the URL with scripts off and count
Prompt baseline	Week 2	Fixed prompts, verbatim answers, dated	Ask the same prompts yourself
Passage rewrites	Weeks 3-5	Before and after, section by section	Read both; the change is obvious
Entity sheet	Weeks 4-6	One description plus contradicting sources	Visit the sources listed
Schema diff	Week 6	Markup claims vs visible copy	Run any schema validator
Monthly measurement	Ongoing	Citation frequency, share of voice, sentiment	Compare against the stored baseline

In scope, out of scope, and the one genuinely blurred line

What an AEO engagement covers, what it does not, and the single boundary that is genuinely ambiguous.

What an AEO service does and does not cover
Area	Included	Why
Crawler access and robots	Yes	The cheapest fix in the discipline
Render diagnosis per template	Yes	The largest cost variable
Render implementation	Usually not	Typically an engineering job for your developers
Passage restructuring	Yes	Retrieval selects passages, not pages
Schema reconciliation	Yes	Contradictory markup is worse than none
Entity consistency	Yes	Stops assistants being wrong about you
Third-party corroboration	Yes	The slow half, and the durable one
Prompt-set measurement	Yes	The only honest way to measure this channel
High-volume article production	No	Volume is not the lever here
Keyword research	No	There is no ranked list to target
Rank tracking as the KPI	No	Cannot see inside an answer
Guaranteed citations	No	Nobody controls model output

How AEO services differ from the SEO services you may already buy

The overlap is larger than most providers admit, and these are the lines where the work genuinely diverges.

Same supplier category, different unit of work
Dimension	SEO service	AEO service
Unit of work	The page and the keyword	The passage and the template
Primary evidence	Rank reports	Server logs, scriptless fetches, stored answers
Main monthly output	Content and links	Structural fixes and a re-run prompt set
Crawlers considered	Googlebot, Bingbot	Plus GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended
JavaScript tolerance	Usually rendered	Frequently not executed at all
Success measure	Position in a list	Named or not named
Time to first signal	Weeks to months	Days for access, months for citations

Service levels, and who each one suits

Three levels of engagement compared on scope, cadence and the kind of business each fits.

Four shapes this engagement takes
Shape	Suits	Contains	Stops when
Audit only	Clean site, cautious buyer	Both tests, baseline, scoped recommendations	On delivery
Technical sprint	Blocked or blind site	Access, rendering spec, verification	Machines can read the site
Full retainer	Question-led category with real debt	All eight artefacts plus monthly measurement	Reviewed at month six
Baseline and monitor	Clean site, brand-led demand	Prompt set recorded and re-run monthly	Competitors start appearing and you do not

Getting found in search

AI, AEO and what is changing

Paid media and lead generation

Websites and design

Choosing and working with an agency

Social, content and brand

By industry and by situation

Frequently asked questions

What are AEO services?

Work that makes your pages retrievable, parseable and quotable by AI answer engines. In practice: crawler access proven from server logs, rendering without JavaScript, passage restructuring, schema reconciliation, entity consistency, third-party corroboration and prompt-set measurement.

What deliverables do I actually receive?

Eight artefacts in the first six weeks: a crawler access report, a robots audit, a render report per template, a dated prompt baseline, passage rewrites, an entity sheet, a schema diff, then monthly measurement against the baseline.

How are AEO services different from SEO services?

The unit of work changes from the page and keyword to the passage and template, and the evidence changes from rank reports to server logs, scriptless fetches and stored verbatim answers.

How much do AEO services cost?

Published US retainers run roughly $1,500-$20,000 per month and a one-off audit $1,000-$4,000. Your figure depends mainly on template count and rendering health.

What is not included?

High-volume article production, keyword research, rank tracking as a success measure, and any guarantee of citation. The blurred line is rendering implementation, which usually belongs with your engineers.

Do you implement the rendering fix yourselves?

We specify it and test it. Actually changing how your application renders is normally an engineering job for your developers, and saying otherwise is how AEO quotes end up wrong by an order of magnitude.

How do I verify the work was done?

Every artefact is independently checkable. Re-run the log query, curl a page with scripts disabled and count the words, ask the baseline prompts yourself, visit the corroboration sources listed.

Why do you store verbatim answers?

Because answers vary between runs. Without stored text there is no way to tell a genuine change from ordinary variance, and a citation count alone cannot be audited by anyone.

How long before I see anything?

Crawler access shows in logs within days. Rendering and structure take weeks. Citation frequency reads honestly from about month six.

Can you guarantee I will be cited by ChatGPT?

No. Nobody controls model output, and any supplier offering that guarantee is describing something that does not exist.

Is there a minimum term?

Six months, because that is the shortest period over which citation frequency says anything reliable. Shorter engagements are audits, not programmes.

Do AEO services include content writing?

Restructuring existing pages, yes. Publishing large volumes of new articles, no — volume is not the lever in this channel.

Which crawlers do you test for?

GPTBot and OAI-SearchBot from OpenAI, ClaudeBot from Anthropic, PerplexityBot, and Google-Extended, which governs Gemini and AI training use.

What if my site is a single-page application?

Then rendering is your dominant variable and we test it before quoting. It may need server-side rendering or prerendering, which is an engineering project rather than an agency line item.

Can I buy just the audit?

Yes, and for many buyers it is the right purchase. It converts assumptions into a scope and sometimes concludes you do not need the rest.

Do you do link building as part of this?

Third-party corroboration, yes — but for agreement rather than link equity. Independent sources repeating what you say is what makes a claim safe to quote.

How is success measured?

Citation frequency, share of voice against competitors, sentiment, prompt coverage, crawler hits and render coverage — all reported separately from search rankings.

Why report AEO separately from rankings?

Because merging them is how weak AEO work hides behind ordinary search gains. If both sit on one chart, one of them is decoration.

Which businesses get the most from AEO services?

B2B SaaS, professional services, healthcare, fintech and manufacturing. Local trades and walk-in retail currently get more from listings and reviews.

What happens in month one?

The crawler access report, the robots audit and the prompt baseline. For a surprising number of sites, month one contains most of the year’s result.

Is llms.txt part of the service?

We will add it, but we will not present it as a strategy. Adoption is low and it is not a ranking factor.

What if the audit says I do not need this?

Then we say so. It happens regularly, most often with clean server-rendered sites in brand-led categories.

Want this done for your site?We build and maintain the search, content and paid programmes described on this page.

Get a free proposal

Get a free marketing proposal

Tell us what you are trying to grow and we will come back with a plan, not a pitch deck. Same-day reply on weekdays.