How long does the free version take?

Around a day to set up the prompt set and rewrite five pages, then an hour a month to sample. That is genuinely the whole commitment at this stage.

When should we start paying for something?

When sampling stops fitting into an hour, when you need more markets than one person can track, or when the diagnosis is clear and nobody has time to act on it.

Does this work for a brand nobody has heard of?

The on-site half does, immediately. The consensus half is slower, because being recommended depends on other people describing you, and that takes longer than editing your own pages.

On no budget

The cheapest way to start

Nearly everything written about this assumes a budget. The version that costs nothing is short, dull and works better than most retainers.

The short answer

The cheapest start is a fixed list of twenty questions your buyers ask, run manually across the assistants once a month and recorded in a spreadsheet, followed by rewriting your five most commercial pages so each opens with a complete answer. That costs time and nothing else, and it is the same first move a paid engagement would make.

What ranked on 23 September 2026: Every first-page result in the scraped set proposed either a tool or a service. The free method is not published anywhere prominent, which is unsurprising given who does the publishing.

What free actually gets you

The unpaid version covers the measurement and most of the on-site work. It does not cover scale, and that is the honest limit.

Manual sampling gives you the same signal a monitoring tool does, for an hour a month, as long as the questions never change
Rewriting for extraction needs judgement rather than software, and you already know your subject better than any vendor
Correcting directory and review descriptions is free and is often the single highest-value change available
What free will not do is cover dozens of markets, thousands of pages, or continuous monitoring
The point at which you should pay is when the manual work stops fitting in the time you have, not when a vendor says so

Things that sound right and are not

ClaimFree means low value.
The expensive parts of this work are labour and attention. The method is not secret, and most of what agencies do first is what is described here.
ClaimWe will start once we have budget approval.
The baseline is what makes the budget case. Starting after approval means arguing for money with no evidence.
ClaimA free tool tier is enough.
Often it is, for a small prompt set. Just keep the raw record yourself, because free tiers change and take their history with them.

Where to start

01Write twenty questions a buyer would ask an assistant about your category, in their words, and freeze the list.
02Run them across the assistants your buyers use, and record whether you appear, what position in the answer, and how you are described.
03Rewrite your five most commercial pages so the first paragraph answers the page's question completely and could be quoted alone.
04Correct how you are described on every directory, review site and profile you can edit.
05Repeat the sampling monthly, changing nothing else for the first two months so you can see what your edits did.

Questions

How long does the free version take?: Around a day to set up the prompt set and rewrite five pages, then an hour a month to sample. That is genuinely the whole commitment at this stage.
When should we start paying for something?: When sampling stops fitting into an hour, when you need more markets than one person can track, or when the diagnosis is clear and nobody has time to act on it.
Does this work for a brand nobody has heard of?: The on-site half does, immediately. The consensus half is slower, because being recommended depends on other people describing you, and that takes longer than editing your own pages.

Doing this yourself and want a second opinion on what the assistants say about you? Tell me what you are seeing, or read the AEO Playbook for the whole framework.