AEO work is sold in three shapes. A one off audit, typically two to eight thousand. A fixed scope project, usually eight to thirty thousand. A monthly retainer, commonly three to fifteen thousand. Those are the ranges we see quoted in the B2B SaaS market, and the spread inside each one is large enough to be useless without knowing what moves it.
What moves it is not seniority of the logo or quality of the deck. It is three specific things, and once you can see them in a proposal the pricing stops looking arbitrary.
A note on where these figures come from, because it matters. They are the ranges we observe quoting against other firms and reviewing proposals clients bring us. There is no industry survey of AEO pricing, and anyone presenting one should be asked about their sample.
Four different jobs wearing one label
Most of the confusion is definitional. When two agencies quote wildly different numbers for AEO, they are frequently quoting for different work and both are being honest.
| The job | What it involves | Typical shape | Observed range |
|---|---|---|---|
| Diagnostic | Prompt testing, log analysis, passage and entity review, a named constraint | One off, two to four weeks | $2k to $8k |
| Technical remediation | Crawler access, rendering, schema, page structure, internal linking | Fixed project, four to ten weeks | $8k to $30k |
| Content and corroboration | Publishing, original research, third party records, digital PR | Retainer, ongoing | $5k to $15k per month |
| Measurement | Repeatable prompt sets across engines, tracking, reporting | Retainer, often bundled | $1.5k to $5k per month |
The second row is where the money usually should go first and rarely does, because it is unglamorous and produces no monthly deliverable to admire. The third row is where most retainers actually sit, and it is the only one of the four that genuinely needs to be ongoing.
The question that reprices most proposals. Ask whether the engagement includes implementation or stops at recommendations. A great many AEO retainers deliver documents and rely on your engineers to act on them, which means you are paying for analysis and still paying internally for the work. That is sometimes correct. It should never be a surprise.
The three things that actually move the number
Who does the work. The difference between a senior practitioner doing the analysis and a junior doing it under review is roughly a factor of three in cost and considerably more than that in outcome on diagnostic work. Both models are legitimate. Only one of them is usually disclosed. Ask who will be in the log files and what their last three engagements looked like.
Whether implementation is included. A recommendations only engagement is cheaper and shifts the expensive part onto your team. If your engineering backlog is eleven weeks deep, a recommendation is not a fix, it is a queue position.
Whether measurement is real. Running a fixed prompt set across four engines, twice each, monthly, and keeping the transcripts is genuine work with a genuine cost. Pulling a vendor dashboard number into a slide is not. Both appear on proposals as reporting. The method is described in our full testing methodology if you want something to compare a proposal against.
Before you evaluate any proposal, it helps to know which of the four jobs you actually need, because that alone changes the budget by an order of magnitude. Our free AI visibility check produces that answer, names one binding constraint, and costs nothing. Take the output to whoever you were going to hire, including us.
The in-house comparison, done properly
Almost every agency comparison is run against the wrong number. People compare a retainer to a salary. The honest comparison is to loaded cost plus ramp time plus the opportunity cost of what that person stops doing.
Published wage data is a reasonable starting point. The US Department of Labor’s occupational profiles put median wages for market research and search marketing roles in a band that, once you add employer taxes, benefits, tooling and management overhead, generally lands somewhere between a quarter and forty percent above the headline salary. The underlying occupational employment statistics let you check the figure for your metro rather than a national average, which for this kind of role varies more than people expect.
| Senior hire, in-house | Retainer, mid-market agency | |
|---|---|---|
| Direct cost, year one | Loaded salary plus recruitment | Twelve monthly invoices |
| Time to first output | Two to four months including notice and ramp | Two to four weeks |
| Breadth of pattern exposure | Your company only | Dozens of companies across categories |
| Continuity risk | One person, who may leave | Account team, which may rotate |
| Institutional memory | Stays with you | Leaves when the contract does |
| Best when | The work is continuous and central to the product | The work is a defined constraint or a ramp |
Marketing budget context is worth having in the room for this conversation. The CMO Survey publishes marketing spend as a share of revenue and of total firm budget twice a year, broken out by sector. It gives you a defensible frame for what proportion of an existing budget this work can reasonably claim, which is a more productive internal argument than debating whether the discipline is real.
Pricing models, and what each one quietly incentivises
Every model creates behaviour. None of them is neutral, and the ones that sound fairest are frequently the worst in practice.
| Model | What it rewards | Where it goes wrong |
|---|---|---|
| Fixed monthly retainer | Predictability for both sides | Drifts toward volume of deliverables once results plateau |
| Fixed scope project | Finishing and leaving | Everything discovered mid-project becomes a change order |
| Hourly or day rate | Transparency about who did what | Punishes the agency for being fast, which is the point of hiring experience |
| Performance based | Alignment, in theory | The metric gets managed. Mentions are cheap to manufacture and hard to audit |
| Retainer plus project | Steady work with defined bursts | Two contracts, two scopes, twice the boundary disputes |
The trade body literature on agency compensation has covered this ground for decades in adjacent disciplines. The Association of National Advertisers has published repeatedly on the drift from commission to fee to incentive models and on how each one changes agency behaviour. None of it is about AEO and all of it applies, because the underlying problem is the same one. You are buying judgement and paying for something easier to count.
The cheapest thing you can do before any of this is buy the diagnosis separately.
A diagnostic is the one purchase here with a bounded cost and an unambiguous deliverable. It also frequently ends with the finding that your existing team can handle the work, which is the outcome nobody selling a retainer is motivated to reach.
Ours is free and you keep the output either way. Book a visibility audit, and put us alongside two other firms if you want a genuine comparison rather than a sales process.
Working out your own ceiling before anyone quotes you
Do this before the first call. It takes ten minutes and it changes how every subsequent conversation goes, because you arrive with a number instead of reacting to one.
Start from what a result is worth rather than from what the work costs. Four inputs.
- Deal value. Your average contract value, annualised. Not your best deal.
- Influenced volume. How many deals per quarter involve a buyer researching the category before they speak to you. For most B2B SaaS this is nearly all of them.
- A plausible shift. What share of those you would expect to move if engines named you consistently where they currently do not. Be pessimistic. One to three percent is a defensible assumption and anyone promising more should be asked to show the working.
- Payback period. How many months of programme cost you are willing to carry before it has to pay for itself. Nine to twelve is normal, and less than six is not realistic for this discipline.
Multiply the first three for a quarterly value, divide by three for a monthly value, then divide by your payback tolerance. That is your ceiling. For a company with a forty thousand dollar contract value and sixty influenced deals a quarter, a two percent shift is worth roughly forty eight thousand a quarter, which supports a meaningful retainer. For a company with a four thousand dollar contract value and twenty influenced deals, the same arithmetic supports an audit and some internal work, and nothing more.
Run it honestly and a good number of companies discover they should not be hiring an agency at all this year. That is a useful finding and it costs nothing to reach. The wider version of that decision, including the cases where in-house genuinely wins, is in the honest comparison of agency, agency and in-house.
Five pricing red flags
A number before a diagnosis. Any firm quoting a retainer before looking at your logs, your pages and what engines currently say about you is pricing a template. The template may be fine. You are still buying a guess.
A guaranteed outcome. Nobody controls whether a model names you. The honest promise is about inputs and about process, which is crawler access fixed, passages restructured, records corrected, measurement run. A guarantee of mentions is either a misunderstanding of the mechanism or a plan to manufacture the metric.
Pricing that scales with your revenue rather than the work. Your funding round does not make the log analysis harder. When a proposal moves because your company got bigger and the scope did not, you are being priced on ability to pay.
A twelve month minimum with no break clause. Six months is a reasonable ask because this work genuinely takes a quarter to show anything. Twelve with no exit is a retention mechanism rather than a methodological necessity.
Deliverables counted rather than described. Four blog posts and a report per month is a production schedule, not a strategy. Ask which constraint each deliverable addresses. If the answer is general rather than specific, the plan is general.
What we would tell you if you called us tomorrow
For most companies under about five million in revenue, the right first purchase is a diagnostic and nothing else. Buy the answer to what is actually broken. A meaningful share of the time it is crawler configuration plus a handful of pages, which is a fortnight of internal work rather than a programme.
A retainer makes sense when the binding constraint is genuinely continuous. That usually means one of three things. You need to publish original material regularly because your category has no independent sources. You need third party corroboration built over quarters. Or you need measurement run properly across many prompts and engines, repeatedly, which is tedious and real work that nobody internally will sustain.
If none of those three describes you, we will say so on the call. It costs us a proposal and saves you a year of paying for the wrong thing, and the companies we say it to tend to come back when the constraint changes.
Related reading
- How to choose an AEO agency. The evaluation criteria that sit underneath the price, including what to ask about method.
- The 20-prompt set to run before you hire. Free, takes an afternoon, and changes what you are buying.
- How Perplexity chooses sources. Useful for judging whether a proposal understands that the engines differ.
- What AEO is. The full picture, if you are still working out whether this is a real discipline or a rebrand.
Work out your ceiling first, then go and collect three quotes.
Ten minutes of arithmetic and an afternoon of prompt testing will tell you more about what you should spend than any pitch will. Arrive at the conversation with a number and a diagnosis and you will find the quality of the conversations improves immediately.
When you want the diagnosis part done for you, book an audit call. It is free, you keep the output, and we will tell you plainly when the answer is that you do not need us yet.
Shaban runs answer engine optimisation for B2B SaaS companies at Uncited Brands. He works the unglamorous end of the problem, mostly crawler access, retrieval diagnostics and measurement baselines, and publishes the tests that do not go his way alongside the ones that do.