GEO: agency, in-house, or software?
Deadpan Labs · August 21, 2026 · 7 min read
GEO · Ecommerce · Buying
Buy software if you have someone in-house who will act on a dashboard every week. A tracking tool will tell you where you are named across the answer engines, and that is worth knowing. What it will not do is change the answer. Measurement is the cheap half of this problem, and the reason most of those dashboards go unopened by month three is that nobody owned the work they pointed at. Do it in-house if your team already produces content and has relationships in your category, because the work that moves AI answers is mostly earning third-party mentions and nobody earns those faster than the people who already know the editors. Hire an agency when the measurement is not the hard part — when you know you are invisible and have nobody whose actual job is to fix it.
Those are three different products solving three different bottlenecks, and picking the wrong one is usually a diagnosis error rather than a budget error.
What each option is genuinely better at
Taken on their merits, and stated the way we would state it if we were not one of the options:
- 1 · Software is better at measurement, breadth, and cost per data point. A tracking platform runs far more prompts, far more often, across more engines than a human process will, and it gives you a time series you own outright. If your question is 'where do we stand and is it moving,' software answers it better than we do. It is also the only option that keeps working at 2am and does not raise its rate.
- 2 · In-house is better at speed, context, and everything relationship-shaped. Your team knows the product, knows which claims are defensible, and can email a niche editor without a briefing document. Since the highest-leverage GEO work is getting included in independent roundups and being genuinely useful in communities, a motivated in-house owner beats an outside vendor on the part that matters most. In-house also compounds: the knowledge stays.
- 3 · An agency is better at execution capacity and at knowing what to ignore. The gap most stores actually have is not information — it is that nobody has time to do the work, and that the list of possible fixes is longer than the list of fixes that matter. An outside team brings hands and a prioritization that comes from having seen the failure patterns before.
- 4 · Software plus in-house is the strongest combination for a store that has the person. Buy the tracking, let your own team act on it, skip the agency entirely. We will say this to your face: if you have that person, that is the better spend.
- 5 · Agency plus software is what most agencies, including us, are actually doing. The tracking is a commodity; the judgement about what to do with it is not. Ask any agency you are evaluating which half you are paying for.
What each one genuinely costs you
Not price — cost. The three fail in different ways, and the failure mode is the part vendors leave out.
- 1 · Software fails when nobody opens it. A tracking dashboard is a diagnosis, not a treatment. It will tell you precisely how invisible you are, every week, forever, and it will not write a single pitch to a roundup editor. The predictable outcome is a subscription and an unchanged position.
- 2 · In-house fails on consistency, not on capability. The panel gets run twice and then a launch happens. Citation outreach gets three emails deep and stops. The work is not difficult; it is repetitive and unglamorous, and it is the first thing dropped when the quarter gets busy.
- 3 · An agency fails when it is bought as a substitute for owning the channel. If nobody internally cares about the answers, an outside team produces reports that get filed. It also costs the most, and it should — you are paying for hours, and the hours are the product.
- 4 · All three fail against the same clock. AI answers move slowly and noisily. Ahrefs analyzed 137,210 domains and found 97% of published llms.txt files received zero requests in May 2026 — a useful reminder that plenty of confidently-sold AI-search tactics do nothing at all. Whichever route you take, ask for the evidence behind each recommendation, ours included.
How to choose in one question
Ask: who at my company will be personally accountable for this number in six months?
If the answer is a named person with time, buy software and let them run it. If the answer is a named person without time, buy software and hire hands for the outreach. If the answer is nobody, an agency is the only option that produces motion — but go in knowing you are buying an owner, not a tool, and price it accordingly.
For what it is worth, here is our position stated plainly. Deadpan Labs runs AI-visibility panels and the citation work behind them for Shopify and DTC stores, and we run the same panels on our own products — one of our brands grew from zero to more than 900 downloads organically with no ad spend, which is the receipt we have. We have no client case studies, because we have not signed clients yet, and we are not going to imply otherwise. That is a real reason to consider one of the other two options, and we would rather you weighed it now than later.
Whichever way you go, start from the same place: the actual answers. Ask the assistants the five questions your buyers ask, write down who gets named, and you will know within an afternoon whether your problem is measurement, execution, or ownership — which is the only thing that decides this. If you want us to run that first read for you, the AI-visibility check is free and comes back with the names and the sources.
The free check
Are the assistants recommending you?
We’ll run your brand against ChatGPT, Perplexity & Google AI and show you who’s named instead.