GEO Services: Agency, Tool, or Do It Yourself?
Measure it yourself, use a tool, or hire an agency: an honest framework for which route fits your situation, plus the question most comparisons skip.

The short answer: doing it yourself costs time but no money, a tool usually costs a fixed subscription fee for repeated measurement across platforms, and an agency costs the most because it adds content work on top of measurement. Which one you need doesn't depend on your budget. It depends on one question most comparisons skip: is your bottleneck that you don't know where you stand, or that you know but have nobody to act on it? That distinction drives the rest of this article.
Three routes, one goal
Whether you measure it yourself, use a tool, or hire an agency, the goal stays the same: find out whether ChatGPT, Gemini, Claude and Perplexity mention your brand when someone asks a buying question in your market, and what to do about it. The three routes only differ in who does the work and how often.
- DIY: you ask the questions, you log the answers, you update your own content.
- A tool: measurement gets automated and repeated; the content work stays with you.
- An agency: both measurement and content work (and often strategy) get outsourced.
None of the three is better by default. They fit different situations, and the biggest risk is picking a route sized for a different budget than your actual bottleneck.
Route 1: do it yourself
The manual route costs no money, only time. Write down ten to twenty buying questions the way a customer would phrase them, for example "which accounting software works for a sole proprietorship," and ask them separately to ChatGPT, Gemini, Claude and Perplexity. Limit personalization where you can (a fresh conversation with no prior chat history, or logged out where that's possible without an account; Claude and Gemini usually require one), since a personalized answer is harder to reproduce. Log for each answer whether your brand is mentioned and in what role. Repeat it a few weeks later with the same questions to tell a pattern apart from coincidence. We describe this approach in more detail in what is AI visibility and how do you actually measure it.
This route works well if you're tracking one brand, have a limited set of questions, and don't mind repeating it by hand. It gets awkward once you want to compare several competitors, cover multiple platforms at once, or risk "we'll do this again next month" quietly not happening because nobody owns it.
Route 2: a tool
A tool takes the repetition off your plate: it pulls fresh answers from the AI assistants on every measurement round, using a question set you freeze, and shows the pattern over time instead of one snapshot. What a tool usually doesn't do is rewrite your content. You see that a competitor gets mentioned where you don't, not automatically why, and certainly not a ready-made fix.
This is the route if you want to scale the measurement work (more questions, more platforms, repeated over time) but want to handle what the data shows yourself, or with your own content team. We wrote a separate eight-question framework for evaluating a tool before you sign in AI visibility tools compared; that framework (does it pull real answers, is the question set frozen, does it avoid ranking guarantees) applies here just as well.
Route 3: hiring an agency
An agency usually adds two things a tool doesn't: content production (rewriting articles, restructuring pages, adding FAQs, clarifying entities) and strategy (which topics, in what order, how this fits your broader marketing). That's worth it when content production is your actual bottleneck: you know what needs to change but have nobody to write it, or no time to tackle it properly alongside everything else.
Hiring an agency just to find out whether you're mentioned is relatively expensive tooling for that question; a tool (or the manual route) already covers it. The reason to pick an agency is almost always the execution that follows, not the measurement itself.
The hidden question: who does the content work?
This is the question most "GEO tool vs. GEO agency" comparisons skip, and it's the most important one. The research that introduced the term generative engine optimization (Aggarwal et al., 2023) tested nine concrete content changes (adding citations, backing claims with statistics, adjusting writing style) and found a visibility gain of up to 40% in their test setup. None of those changes is something a measurement tool does for you automatically; it's work on the content itself. A tool shows you where you stand. Whoever actually changes the content determines whether that shifts. With DIY and with a tool, that's you. With an agency, that's (part of) what you're paying for.
Cost versus what you get
| Route | What you pay | What you get | What you don't get |
|---|---|---|---|
| DIY | Time, no money | Full control, low barrier to start | Doesn't scale well across brands/platforms |
| Tool | Usually a fixed monthly subscription | Automated, repeatable measurement | Content production and strategy |
| Agency | The most, usually a monthly fee | Measurement plus content work and strategy | Guarantees (which don't exist anyway) |
The exact level of these numbers can't be pinned down here: tool pricing varies with the number of platforms and measurement frequency, and agency pricing varies even more by scope, with many agencies not publishing rates at all. Don't expect a fixed figure; ask specifically what's included in the scope behind whatever number you're quoted.
Red flags, across all three routes
Whether you're evaluating a tool or reading an agency proposal, the same red flags apply:
- A ranking guarantee, or a promise that you'll come out "on top" in AI answers. That doesn't exist. AI answers are non-deterministic; OpenAI itself describes its own models' output that way explicitly (developers.openai.com).
- A "potential" score that claims how well you could score without any actual mentions having been measured. That's an estimate dressed up as a measurement.
- No visibility into the method: which questions, which platforms, how often it measures. A vendor, tool or agency, that can't show you that is asking for blind trust in a number you may end up reporting internally.
An example (illustrative)
Say a B2B software company notices competitors get mentioned more often in ChatGPT answers to buying questions in their market. Three scenarios, same starting point. Scenario A: the marketing team asks twenty buying questions themselves across the four big assistants, finds their content is light on concrete numbers and citations, and fixes that themselves in the existing articles. Scenario B: they set up a tool to repeat that monthly across twenty questions and three competitors, and keep updating the content themselves based on what the tool shows. Scenario C: they hire an agency that both measures and delivers a content calendar of rewritten articles, because the in-house team simply has no time for the rewrite work. No scenario is wrong; they fit different bottlenecks. This example is illustrative and doesn't describe a specific real company.
Can you combine them?
Yes, and in practice that's often the normal sequence: measure it yourself first, briefly, to see whether there's a problem; add a tool for repeated, scaled measurement; bring in an agency once content production becomes the bottleneck. There's no reason to start with the most expensive option before you've even answered whether there's something to improve.
Where we fit into this list
To be upfront: GEO-Crafter is a tool, not an agency. We ask the buying questions your customers actually use, in parallel, to ChatGPT, Gemini, Claude and Perplexity, with a question set you freeze, and we classify who counts as a real competitor before we calculate a score. No ranking guarantee, no potential score: a coverage percentage based on real answers, repeatable over time. The content work that follows from that measurement stays with you, or with an agency, whichever fits your situation. Curious where you stand right now? Request access or take a look at the GEO audit first.
Frequently asked questions
Should I measure this myself before hiring a tool or agency? It helps. A few weeks of manual measurement shows you whether there is actually a problem (are you already mentioned, and in what role) before you spend money on a fix. Without that baseline, you won't know later whether a tool or agency changed anything, or whether you were already doing fine.
Can a generative engine optimization agency guarantee I get mentioned more often by ChatGPT? No, and an agency that puts that in a contract deserves extra scrutiny. AI answers are non-deterministic: the same question can return a different result on the next call. An agency can improve content and structure, but a guaranteed mention is not something anyone can technically promise.
What's the difference between a GEO tool and a GEO agency? A tool measures: it repeatedly pulls real answers from the AI assistants and shows whether you're mentioned. An agency usually adds the content work on top: rewriting articles, restructuring pages, clarifying entities. Some agencies use a tool as part of their own service; others just deliver advice based on a one-off analysis.
When does doing it yourself stop making sense? Once you want to repeat it across multiple platforms with a fixed question set, and copying and logging answers by hand costs more time than it's worth. For one brand and a handful of questions per quarter, manual is fine. Once you add more brands, competitors and platforms, it quickly becomes a spreadsheet nobody keeps up with.