Hiring help · Oct 6, 2026 · By Gerald De Oleo

AEO In-House or Outsourced? The Half Your Team Won't Finish

The short answer: if you have someone on content with real hours and the patience for six months, you can do half of this in-house and it will work. The other half is the half that decides whether you appear at all — correcting the third-party sources that describe you wrong. It isn’t hard, it’s tedious, and it’s never Monday’s priority. Before you decide either way, buy a diagnosis first, like an AI visibility audit. It’s the cheapest thing in this category and it tells you which half is actually your problem.

I sell this work, so I have an obvious stake here. Which is why I’m going to be specific about when you shouldn’t hire me.

Split the work before you decide

“In-house or outsourced” is the wrong question, because this isn’t one task. It’s four, and they differ enormously in difficulty.

1. Diagnosis. Running your buyer’s real questions across every engine and recording what each one says about you. Difficulty: low to execute, high to design. Anyone can open ChatGPT. Few people know what to ask, because the questions that matter carry the buyer’s stack, team size and budget attached — and they aren’t the questions you’d guess.

2. The technical layer. Making sure AI crawlers can read you without JavaScript. Difficulty: low if you have a developer. Do this in-house, always. Don’t pay anyone for it. Two minutes with the Crawler Check settles it — paste a URL, see exactly what each bot receives. If it comes back empty, stop everything else: nothing you publish exists to the models until that’s fixed.

3. The pages that decide the answer. Comparisons, integrations, segment pages, pricing and proof. Difficulty: medium. The hard part isn’t writing, it’s specificity. Your team will ship the generic page, because it’s easier and because nobody internally wants to be the one who decides which segment you’re talking to.

4. Correcting how the web describes you. Directory categories, old roundups, third-party comparisons, outdated mentions. Difficulty: low per task, high in aggregate. It’s tracing, cold email and follow-up over several weeks.

Look at that list again. What an internal team does well is 2 and 3. What it doesn’t finish is 4. And 3 without 4 doesn’t compound — you publish on top of a broken record, and the model files the new material under the category it already settled on.

That’s the most expensive mistake in this category: buying content before fixing the entity. If the web describes your company three different ways, more content amplifies the confusion.

“What if I only outsource part 4?”

That’s legitimate and sometimes what I’d recommend. But 4 without 1 is blind — you don’t know which sources to correct until you know which ones the model is reading. The minimum viable combination is 1 + 4, and that can be bought on its own.

When in-house actually works

I’d rather say this plainly than take six months of your money for nothing.

If your category is young and uncontested. Ask about your niche in ChatGPT. If it answers in generalities and doesn’t name anyone specific, there’s no record to correct and no incumbent to displace. Publish with judgment and you’ll win on ordinary effort.

If your description on the web is already accurate. If you never repositioned, never renamed, don’t run two domains, and what people wrote about you in 2023 is still true — part 4 barely exists for you. What’s left is content, and that’s your team’s job.

If you have someone genuinely assigned. Not someone giving it Thursday afternoons. Someone with this in their quarterly objectives, able to hold the pace for six months without another priority taking it.

In those three cases, hiring me is slow and expensive compared to what you’d achieve alone.

And there’s a fourth: if you already have an SEO agency, the market adds roughly 20–30% to an existing retainer for this work. It’s usually the cheapest route — assuming your agency can actually do it, which is a question I’ll come back to.

Three signals in-house won’t get there

You already tried for three months and nothing moved. Almost always because you attacked 3 without doing 1. You published good content that the model filed under the wrong category. The content wasn’t the problem; the order was.

Your company changed and the web didn’t notice. Repositioning, new category, new product, a merger, two domains. Here part 4 is 80% of the work, it’s the most tedious, and it’s what a team with twelve other priorities never finishes. Old descriptions don’t expire on their own — they compete with the new one until someone corrects them.

You can verify this signal yourself before reading further. Models learn from snapshots of the web, and those snapshots have dates. Paste your domain into the Positioning Check to see how your homepage described you over the last three years, and which version the recent crawls captured. If the version on record isn’t the company you sell today, you have your answer and there’s no further debate about doing this in-house.

You compete against an incumbent who wrote the guides. When your category’s comparison content was written by your competitor, or by someone they sponsored, you don’t get out of that by publishing on your own blog. In attribution software, six of the nine results I checked were published by companies selling attribution software. That doesn’t get displaced from the inside.

What each path costs

Outsourced, 2026 market figures:

There’s a floor on full retainers, and most of the category agrees on it: below roughly $1,500–$2,000 per month, what’s usually being sold is traditional SEO with a new label. If you’re quoted less for a full program, ask which deliverable is specifically different from an SEO one. A standalone measurement subscription is a different product and sits below that floor legitimately — it isn’t the work, it’s the scoreboard.

For reference, my own audit is fixed at $499, and ongoing work after it is $2,500 a month with no contract. The numbers are on the services page.

In-house: half a person for six months, plus whatever it costs to discover what doesn’t work by trial and error. Don’t compare these on price. Compare them on whether they finish. A cheap internal path that never completes part 4 costs more than outsourcing, because you lose the six months anyway and still don’t have the diagnosis.

“Can’t I just start with a monitoring tool?”

Published tool pricing runs from $29 to $489 per month, and they do one thing: show you the symptom every week. If you already know your positioning is clean and you have people who can act on findings, a tool plus your existing team may be enough.

If you don’t know why you’re missing from the answers, a dashboard will show you that symptom indefinitely without ever naming the cause. A $200 dashboard and a $6,000 retainer aren’t competing offers — they answer different questions.

The mixed approach

For most situations, the thing that makes sense isn’t all in or all out:

• I take 1 and 4. The diagnosis built from your buyer’s real questions, and the correction of sources you don’t control.
• Your team takes 2 and 3. The technical layer and the pages. They already know how, and they know the product better than I do.
• We measure together, monthly. Same questions, same engines, so there’s a before and after you can actually show.

It costs less than handing over the whole thing, your team keeps the judgment, and the part that never gets finished on its own gets covered.

The honest test for deciding: ask what proportion of the fee goes to thinking versus producing. If what you need is production at volume, buy production. If what you need is to know what’s broken, buy that — and don’t pay agency overhead for it.

What to expect in the first 90 days

This works as a yardstick either way — for measuring a vendor or for measuring your own team.

Days 1–30. A baseline: your prompt set, per engine, with recommendation status and the exact description each one gives. An entity inventory — every listing, profile and comparison page describing you, checked against how you describe yourself today. A prioritized list of corrections.

Days 31–60. Corrections shipped. Fixed categories, schema, integration and pricing pages, the gaps your buyers’ constrained questions require. First re-measurement on the same prompt set.

Days 61–90. Movement on engines that search the web on every query, where your own pages carry the most weight. Third-party work underway but probably not reflected yet — that’s the slow part.

What you shouldn’t expect is a number for the board by day 30. What you should demand is knowing, by day 30, exactly why you’re missing from the answers that matter.

Quick answers

Can my SEO agency do this? They can do 2 and part of 3, because it resembles work they already do. Part 4 isn’t SEO. Ask your agency the same question you’d ask any new vendor: what exactly do you measure, and how do you verify the description is accurate? The answer tells you whether they built real capability or relabelled existing deliverables. More on what carries over and what doesn’t in AEO vs SEO.

Do I need a full-time hire? Almost never. The load is heavy for the first two months and drops considerably after. It’s project-shaped, not headcount-shaped.

How long until something shows? Changes to your own site register sooner, in engines that search the web on every question. Third-party changes take longer, because they depend on other sites updating and being re-crawled. Anyone promising a date for a specific engine is guessing.

What happens if I do nothing? If your category is already contested in AI answers, the record that exists about you today is the one that keeps getting cited. It doesn’t correct itself.

What’s the most expensive mistake? Buying content before fixing the entity. The correct order is in How to Get AI Recommendations.

If I do decide to hire, how do I choose? Eight options with their published prices are compared side by side in Best AEO Agencies for B2B SaaS. The ten questions, the red flags and the full price bands are in How to Choose an AEO Agency, and what a proposal should contain is in What to Ask an AEO Agency Before You Sign.

How I work

I’m a solo operator — a one-person AEO agency for vertical B2B SaaS — which fits when the problem is diagnostic rather than volume: entity confusion, a repositioning that didn’t propagate, a strong product with an unclear category. It fits badly if what you need is sustained content production across many pages, heavy PR and placement, or several markets at once. In that case an agency serves you better, and I’ll tell you so.

The diagnosis comes before the proposal, not after. I run your buyer’s questions across ChatGPT, Perplexity, Gemini and Google before we speak, so the first call opens with findings instead of a pitch: where you make the shortlist, how you’re described, and who takes your place when you don’t.

From there the work runs in order — fix what describes you wrong, build the pages that answer the constrained questions, measure the same questions every month. Full scope is on the services page.

You don’t have to decide this blind, and you don’t have to start with a call. Two free checks, no conversation required:

• Crawler Check — what each AI bot actually receives when it visits your site. Two minutes.
• Positioning Check — which version of your company got captured in the snapshots the models read. Forty seconds.

If both come back clean, your team can probably carry this and I’ve just saved you a sales conversation. If either one doesn’t, send me your domain and, before we talk, I’ll run a free Shortlist Check: three or four of your buyers’ questions in one engine, to see whether you appear and how you’re described. Book a call.