GEO In-House vs Agency vs Tools: The Total-Cost Model

"Should we do this in-house, hire an agency, or just buy a tool?" is really three questions wearing one budget line, and the market profits from the confusion. In-house and agency are two ways to buy the same production labor. Tools are a third product entirely — measurement — that produces no pages at all. We run the in-house version across 3 production brands and 400+ fact-audited pages, and we sell builds; both facts shape this page, so we show the arithmetic and you can check it [our data].

What are the three ways to buy GEO work?

Two of the three buy production; one buys reporting. That asymmetry is the single most common budgeting error we see, so it goes first:

The work itself is not mysterious. The Princeton GEO study's best-performing tactics were content-level changes — adding citations, quotations, and statistics to pages (arXiv 2311.09735, KDD 2024; a benchmark result, not a field guarantee) — and that is drafting and evidence work under any label, the practice documented across the operator's guide.

What does in-house GEO actually cost?

Your fully loaded hourly rate multiplied through 5 stages per page: research, drafting, fact audit, publish QA, and scheduled refresh. Our illustrative model — the labor mix is how our pipeline runs; the hours and wages are yours to replace — prices a fact-audited page at 7 lifetime hours, worked through in what a fact-audited page costs [our data].

The multiplication for a 100-page first phase:

Two things the table understates. The one-time engine work — the content contract, the build gates, the schema templates — is what makes page 100 cost the same discipline as page 1; on our fleet that setup is measured in days, not months, because it is the same engine each time [our data]. And refresh is inside the 7 hours for a reason: content under 3 months old was about 3x more likely to be cited in ChatGPT answers in Kevin Indig's 2026 analysis (Growth Memo, 2026) — a correlation, but one that makes "publish and forget" a plan to decay.

What does a GEO agency actually cost?

There is no standard number, because the label carries no standard unit — the full argument, including the red flags, is in how agency pricing actually works. What a retainer buys is deliverable volume: pages produced, technical tickets closed, reports shipped. So the only honest comparison is cost per deliverable, checked against your own wage math.

The in-house model above prices a fact-audited page at $350 to $700 of labor at $50 to $100 rates. An agency page quoted meaningfully below that range is not a bargain — it is a different product, usually missing the audit and refresh stages, which are about 43% of lifetime labor in our illustrative model. An agency page quoted above it should come with something wages alone don't buy: a working engine, committed audit ledgers, and measurement you keep when you leave.

What do tools cover — and what don't they?

Tools measure; they do not produce. The category spanned $29 a month (Otterly) to over $1,000 (Profound) on vendor pricing pages in August 2026 — $348 to $12,000+ a year, priced in the tool pricing comparison — and every dollar of it buys sampled visibility reporting, not a single page of the library being measured.

That makes the tool line optional on top of either production path, not a third way to get the work done. Our own answer is to skip it: all 3 of our production brands run on free measurement — Search Console mining, a hand-run prompt battery, and on our insurance build a GA4 AI-referral channel — with no paid visibility tool anywhere in the fleet [our data]. Reporting obligations at agency or enterprise scale change that math; curiosity does not.

When is an agency — including ours — the wrong buy?

More often than sellers admit. If you have a content team, the evidence-backed core of this work is learnable structure — answer-first pages, extractable passages, sourced claims — not proprietary technique. If your niche needs 20 pages, the engine overhead an agency amortizes across 150 pages is overhead you are renting for nothing. If niche fit is unproven, spending a content budget to discover an unwinnable market is the expensive version of a question a research pass answers cheaply.

And if the pitch promises citations or AI traffic, walk at any price: answer engines decide what they cite, and the click side is modest even when they do — Pew measured users clicking a cited source in roughly 1% of visits where an AI summary appeared (Pew, March 2025 data). The honest case for paying anyone, us included, is capacity and process: dozens of audited pages, gates that hold at velocity, measurement wired in from day one — bought as assets you keep, not activity you rent.

How do you choose for your situation?

Price all three paths over 12 months against the same page count, and refuse any comparison that hides a stage. In-house: wages through the 5-stage model, plus the one-time engine. Agency: the quote, decomposed to cost per deliverable, with audit standards and exit terms in writing. Tools: a reporting line you add only when someone must consume the reports.

The pattern in our own fleet is the tell: the money is almost entirely labor, the infrastructure rounds to zero, and the audit stage is the part nobody discounts honestly [our data]. If the budget conversation is really a "should we build at all" conversation, run the fit check first — it is the cheapest line item on this page.