There are three ways to get generative engine optimization done: your own team does it, a monitoring tool watches it, or an agency runs it. They are frequently compared as if they were three prices for one thing. They are not — they buy different things, and two of them are cheaper than us for good reasons.
| do it yourself | monitoring tool | agency | |
|---|---|---|---|
| measurement | Manual, and usually stops after the second month. | Its core strength. Continuous, at a scale humans cannot match. | A fixed panel re-run every cycle, with the answers logged verbatim. |
| deciding what to fix first | Yours to work out. | Not included. A dashboard reports; it does not sequence. | The main thing you are paying for. |
| implementation | Your developers and writers. | None. | Ours, your team’s, or split — and the split changes the price. |
| getting into cited sources | Possible but slow; it is relationship work. | Will tell you which sources matter. Cannot get you into them. | Included, and the slowest part of the loop. |
| typical cost | Existing salaries, plus the risk it stalls. | A monthly subscription, well under a retainer. | From $2,000/mo. |
| fails when | It is nobody’s job and the sprint gets busy. | Nobody acts on the dashboard. | You had the capacity in-house all along. |
when doing it yourself is the right call
One market, one domain, a developer with capacity, and someone who writes. Most of the technical layer is documented, unglamorous work — crawler access for the user-agents engines actually send, a response that does not depend on JavaScript, structured data that describes you as an entity, an llms.txt that maps the site rather than dumping the blog, named authors and a verifiable about page. We put the whole method in our 100-audit report precisely so this is possible.
The thing that kills in-house attempts is not difficulty. It is that measurement has no owner. The work is front-loaded and the proof arrives months later, so the panel stops being re-run around week six and nobody can say whether any of it worked.
when a monitoring tool is the right call
There is now a category of tools — Profound and Peec among them, alongside AI-visibility modules inside the established SEO suites — that track how brands appear across assistants. If your constraint is visibility rather than capacity, they are good value: they watch far more prompts, far more often, than any manual panel, and they surface which sources are being cited.
What they do not do is decide or implement. A tool that tells you that you appear in nine percent of answers has told you a fact, not a plan. If you have people who can turn that fact into a sequenced backlog and ship it, a tool plus your team is a genuinely cheaper setup than hiring us, and we would rather say that than sell you a retainer you do not need.
when an agency is the right call
When the constraint is that nobody owns it. Also when the situation is structurally messy in ways a dashboard cannot untangle: several brands or mirror domains that need to resolve into one identity, multiple markets each needing their own panel and baseline, or a site an engine currently cannot fetch at all. And when the citation layer matters, because getting into the sources engines already trust is relationship and publishing work that no subscription performs for you.
The honest limit on our side: we cannot promise a position in a model’s output, and neither can anyone else. What we run is a loop with a baseline at both ends, so the effect of the work is visible — including the cycles where a prompt did not move.
the part all three share
Whichever you pick, the first step is identical: find out what the engines say about you right now, in your category, with sources. Without that baseline, do-it-yourself has no priorities, a tool has nothing to compare against, and an agency is guessing. That is why our audit is manual and complimentary, and why companies who take it and then do the work themselves are a normal outcome rather than a lost sale.