What you can do manually, for free
More than you might expect.
- Check crawler access. Read the client’s robots.txt and look for each AI user-agent by name. Slow, but free and definitive.
- Ask the questions. Open each engine and ask the buyer questions yourself. Record what comes back and which sources are cited.
- Read the winners. When a competitor page is cited, open it. Compare structure, directness and how quotable the passages are against the client’s page.
- Rewrite for answerability. Nothing stops you restructuring a page so the question is answered directly, high up, in a passage that can be lifted whole.
For one client and a handful of questions, that is a legitimate way to work. Do not let anyone tell you a tool is mandatory.
Where it breaks down
Three places, and they arrive quickly.
Repetition
AI answers vary between runs. Asking once tells you what happened once. To know whether a result is stable you need several passes, and doing that by hand across six engines is where the hours disappear — ten questions across six engines with three passes is 180 separate lookups.
Recording it consistently
Manual notes drift. Two months later you cannot reliably compare because you did not capture things the same way both times, and the comparison is the whole point of the exercise.
Proving it to a client
A spreadsheet you assembled by hand is a weaker artefact than a report with the evidence attached. It is not that the work is worse; it is that it is harder to trust, and trust is what renews the engagement.
The real question is not tool or no tool
It is whether the cost fits the shape of the work.
Most tools in this category bill monthly. That is the right shape if AI visibility is a standing job with alerts and dashboards. It is poor value if the work happens in bursts — per client, per launch, per quarter — because you pay through the gaps.
Pay-per-use inverts that: the cost lands inside the project you are already invoicing. A ten-prompt run across six engines is 15 credits, a URL audit 5, a rewrite 10. $1 = 1 credit, and a quiet month costs nothing.
At those rates the arithmetic is simple. If a manual pass takes you three hours and a tooled pass costs about $85 in credits, the tool is cheaper the moment your time is worth more than $30 an hour.
An honest answer about when a subscription is better
If you have a retained client base large and stable enough that a flat monthly fee works out below per-job pricing, buy the subscription. At our rates that is roughly six or more full site passes a month.
If your clients need their own logins and live dashboards, several platforms ship that and we do not. And if you need white-label reporting today, that is worth checking before you commit either way.
We would rather you bought the right shape of tool than bought ours. The question is whether your work is continuous or project-based, and only you know that.
Common questions
Can I really do this manually?
For one client and a handful of questions, yes. It breaks down on repetition - AI answers vary between runs, so a single check is an anecdote - and on producing an artefact a client can trust.
What is the cheapest way to start?
A free Site Check on the client's domain, which needs no card. Then a ten-prompt run across six engines is 15 credits, and every account starts with 15 free.
Do I need a subscription to offer this as a service?
No. Pay-per-use means the cost sits inside the project you are already invoicing, and a quiet month costs nothing. A subscription only wins if your volume is high and steady.
How many hours does a manual pass take?
It varies with scope, but ten questions across six engines with three passes is 180 lookups before you have written anything down. That is usually where people decide.