Two offers are sitting on the table. One is a hiring file: what a senior digital/AI specialist expects in salary, the benefits, the equipment list. The other is an agency proposal: a monthly retainer, scope clauses, a three-month minimum commitment. Put the two numbers side by side and they usually land dangerously close together, and the meeting collapses into a single question: which one is cheaper?
That is the wrong question, because the two offers are not selling the same thing. The hiring file sells you one person's time; the agency proposal sells you a specific output from a team. The right question is this: which part of GEO work has to stay inside the company, and which part gets expensive if you keep it inside? The manager who answers that one does not say "we picked the wrong model" in month three.
GEO is not one job, it is five separate jobs
This is the step most often skipped. "Let's get GEO done" sounds like a single service, but it contains five jobs that are independent of one another, and each has a different answer to "inside or outside".
| Piece | What it covers | Where it naturally belongs |
|---|---|---|
| Technical foundation | JSON-LD schema, robots.txt and llms.txt, page structure, server-side rendering, speed | Can go outside; it is project work, not continuous |
| Content production | Topic selection, writing, internal linking, updating | Mixed; the skeleton from outside, the accuracy from inside |
| Measurement and tracking | Visibility tracking across AI engines, log/bot analysis, dashboards | Can go outside, but data ownership stays inside |
| Independent sources and reputation | Third-party mentions, directories, reviews, trade publications | Mostly outside; relationship management inside |
| Reporting and interpretation | Deciding what is working, reallocating the budget | Has to stay inside |
A decision taken without making this split compresses five separate decisions into one signature. If you want to see how the pieces connect to each other, our step-by-step guide to building a GEO strategy lays out the order; the decision here is who each step in that order gets handed to.
The pieces also have different time profiles. The technical foundation demands a heavy push at the start and then drops to maintenance — meaning it does not keep a full-time hire busy. Content production is the opposite: it is never finished, it starts again every month. That single difference explains why "one person will do all of it" keeps jamming. Whoever spends the first three months on infrastructure produces no content; by the time they start producing content, what they learned about the infrastructure sits unused.
Four things you cannot outsource
The places where the agency model breaks down are nearly always the same four. They have nothing to do with the agency being bad at its job; this knowledge sits inside the company by its nature.
Product and service knowledge. An agency cannot guess your price tiers, why you turn certain work down, or which customer profile the job goes wrong with. Unless it is written down, that knowledge does not travel outward. And this is exactly the level of detail AI engines quote; generic description does not get quoted.
Price and scope approval. Every price, timeline and commitment that appears in content is a commercial promise. The authority to approve one is never handed outside, under any circumstances. Without an approval mechanism the agency will either write something vague or write something wrong, and both cost you.
Customer data. Reviews, support tickets and sales call notes are the most valuable raw material GEO has, because they show how the real questions get asked. But that data contains personal data. If it is going to leave the building, the contract has to state which fields are anonymised, where the data is processed and how long it is kept.
Sector-specific accuracy review. In fields like health, law and finance, whether a sentence complies with regulation is signed off by the responsible person at the institution, not by the agency. A fast content pipeline that skips this review step gives back the entire investment at the first complaint.
If those four stay inside, everything else is negotiable.
The real bill for an in-house team is more than the salary line
The number in the hiring file is only one line of the real cost. In bands cited by industry sources for the Turkish market, the total annual cost of a senior AI/digital specialist is expressed in millions of lira — but that band varies considerably from business to business, by city, by seniority and by working arrangement. Working the number out in your own spreadsheet is far more useful than borrowing somebody else's average. The line items to include:
- Gross salary and employer contributions
- Benefits: private health cover, meals, transport, bonuses
- Equipment and software licences
- Tool subscriptions. In GEO, visibility tracking, log analysis and content tools are each paid for separately; which tool you actually need is something our comparison of GEO tools can help you decide
- Time to hire. Posting, interviews, offer and notice period add up to months more often than not. The work stops during those months; your competitor does not
- The learning curve. Even a good digital specialist spends their first quarter learning your sector and how AI engines behave
- Management time. Briefing this person, prioritising their work and approving it eats into a manager's week
- Key-person risk. This line never shows up in the table and it is the most expensive one: when the person leaves, the knowledge leaves with them. Schema logic, measurement setup, which experiment failed and why — if none of it is written down, you start from zero
There are also cases where an in-house team is the right answer, and it is only honest to say so. If the product is complex, if content volume is high and continuous, if the sector is hard to learn from the outside, or if the technical capability already exists inside, an in-house team works both faster and deeper. What an agency has to learn from a brief, the in-house team already knows. There is a scale effect too: if the same person produces three pieces a month the cost looks high, but at ten pieces a month the unit cost drops below what you would pay outside. The economics of an in-house team are built on volume.
The agency's real bill is more than the retainer either
The same honesty applies on the agency side. Industry sources put monthly retainer bands starting in the tens of thousands of lira and running up into the hundreds of thousands, with project-based work priced separately. That is a band as well, not a fixed price. The items hiding underneath the retainer:
- Extra invoices for out-of-scope work. Any request beyond the monthly output count written into the contract gets priced separately. If the scope clause is vague, this line turns into a surprise
- Internal time spent writing briefs. However good the agency is, it gets the raw material from you. Weekly calls, gathering sources, approval rounds — all of it is your team's hours
- Approval-cycle delays. If content approval is waiting on you the agency cannot work, but the retainer keeps running
- Handover and learning time. The first few weeks go on the agency learning you, and that time is billed
We wrote up how the price bands form and what usually goes inside a package in our guide to GEO pricing and packages. For the concrete criteria to apply when comparing proposals, look at the 12-question list for choosing a GEO agency.
The only method that makes the comparison meaningful is to reduce both sides to one formula: total monthly cost of ownership ÷ the number of pieces that actually went live that month. Comparing a salary directly against a retainer is adding apples to pears.
The decision matrix: five variables
Place your own situation against the five variables below. No single row decides it; look at which column the weight collects in.
| Variable | Points to an in-house team | Points to an agency | Points to a hybrid |
|---|---|---|---|
| Monthly content volume | 8+ pieces a month, continuous flow | 2-4 pieces a month | 4-8 pieces a month |
| Technical debt | Foundation is sound, the team already maintains it | No schema, rendering broken, speed poor | Foundation gets fixed once, then maintained |
| Regulatory load | High: every sentence needs internal approval | Low: approval is quick and simple | High, but so is production volume |
| Urgency | You have time and can invest over 6-12 months | You need output within 4-8 weeks | Fast start plus long-term transfer |
| Existing internal resource | Someone who will own the topic already exists | Nobody, and nobody is coming soon | An owner exists but has no time to execute |
The three readings the matrix produces most often in practice:
If three or more of your rows collect in the in-house column, hiring makes sense — but set aside part of the budget for outside support so the person you hire is not left alone. If three or more rows sit in the agency column and nobody inside is taking ownership, starting outside before hiring reduces the risk: a bad hire is more expensive than a bad agency. If the rows are scattered, which is the most common picture, the answer is hybrid.
The hybrid model: one owner inside, an execution team outside
In practice the arrangement that works most often is neither pure in-house nor pure agency. One owner sits inside; the execution team works outside.
The owner's job is not writing. The owner's job is this: transfer product and service knowledge outward, give price and scope approval, review sector-specific accuracy, set the priority, and interpret the report and reallocate the budget. In other words, all four items on the "cannot be outsourced" list sit with this person. In most companies the role is not full-time; it runs on a few hours a week from the marketing lead or the founder.
The reason this works is simple: the expensive, repeating part of GEO (schema setup, tracking infrastructure, the content production line) is scale work, and it is done cheaper and faster outside. The cheap but critical part (accuracy, approval, priority) already exists inside the company — it just has not been given a name. The hybrid model puts each of the two where it belongs.
The second benefit of the hybrid is that it spreads the risk. In a pure in-house model all the knowledge collects in one person, and when that person leaves the work stops. In a pure agency model the knowledge is entirely outside, and the same gap opens when the contract ends. In a hybrid, the critical knowledge — what was done, why it was done, what comes next — stays with the owner inside, while the execution side is replaceable. That is the only structure that makes changing model twelve months later cheap.
It has one condition: the owner has to be genuinely appointed. In companies that say "everyone looks after it a bit", the agency model and the in-house model collapse in exactly the same way. If the owner is undefined the agency does not know who to ask and the in-house team does not know what to prioritise; in both cases the invoice runs and the output does not.
If you change your mind twelve months from now
Moving in either direction is possible, but it is not free, and it gets expensive if nobody thinks about it while the contract is being written.
Moving from an agency to in-house, here is what you need to take over: analytics and search console accounts (they should have been opened in your name), tracking dashboards and the logic behind the setup, schema code, the content calendar and archive, and the decision notes explaining why each experiment was run. Without these the new team starts from zero and a year of learning is erased.
Moving from in-house to an agency, what usually gets lost is context: the knowledge in the departing person's head was never written down anywhere. The antidote is documenting the processes even while you are running an in-house model.
So whichever model you choose, put these clauses in the contract or the internal policy: all accounts opened in the company's name, ownership of produced content and code belonging to the company, at least a two-week structured handover at the end of the contract, and a line-by-line statement of what that handover covers. Those four clauses cut the cost of the transition twelve months from now from a few weeks down to a few days.
The short answer
What has to stay inside is clear: product knowledge, price approval, customer data, sector-specific accuracy review and the interpretation of the report. None of that requires a full-time hire; it requires a named owner. The rest — technical foundation, production line, measurement setup, independent-source work — can be done inside or outside depending on your volume, your urgency and your technical debt.
Once you have your own spreadsheet, if you want to talk through where each piece should sit, get in touch; we will look at your current setup and work out together which model loses you less at your volume.
Frequently Asked Questions
Is an agency or an in-house team cheaper for GEO?
There is no directly comparable "cheaper" answer, because the two models are not selling the same thing: a hire prices one person's time, an agency prices a defined output. The meaningful comparison is to divide each side's total monthly cost of ownership by the number of pieces that actually went live that month; on the in-house side you add benefits, tool subscriptions, time to hire and the learning curve to the salary, and on the agency side you add out-of-scope invoices and the internal time spent writing briefs to the retainer. Any comparison made without working out those two numbers is misleading.
Which part of GEO work definitely cannot be outsourced?
Four items have to stay inside the company: the product and service knowledge itself, price and scope approval, access to customer data, and sector-specific accuracy review. Add to those the interpretation of the report, meaning the decision about where the budget moves next. An agency cannot guess this information; unless it is handed over in writing it will produce vague or incorrect content, and that comes back as a correction cost.
Should a small business hire someone full-time for GEO?
If your monthly content volume is consistently below eight pieces and your infrastructure can be fixed once and then run on maintenance, a full-time hire will sit idle most of the time. The more efficient arrangement is to name an owner inside who gives it a few hours a week and send the execution outside. A full-time hire pays for itself when volume is high and continuous, when the sector is hard to learn from the outside, or when regulation means every sentence has to be approved internally.
What does the internal "owner" do in a hybrid model?
The owner is not the person writing content, it is the person deciding. The duties: transfer product and service knowledge to the outside team, approve every statement involving price or scope, review sector-specific accuracy, decide what gets done first, and interpret the incoming report to reallocate the budget. In most companies this role is not full-time and is usually taken on by the marketing lead or the founder; what matters is not the person's job title but that the role has genuinely been assigned to someone.
What should I take over when moving from an agency to an in-house team?
The handover list has to include: analytics and search console accounts (opened in the company's name from the start), visibility tracking dashboards and the logic behind the setup, the schema code used on the site, the content calendar and the full content archive, plus the decision notes explaining which experiment was run, why, and what came of it. If these items are not written into the contract at the outset, a year of learning disappears during the transition and the new team has to start from zero.