Month by month, with the term, the exit and the case for not signing one.
An AEO retainer buys continuous measurement, content built to be quoted, and entity work that compounds. Published rates run roughly $1,500 to $12,000 a month; ours are £2,000, £4,250 and £7,750, on a three-month minimum.
The question behind “what does a retainer cost” is usually “what am I paying for in month four when the obvious fixes are done”. That is the right question, and most pricing pages avoid it.
| Period | The work | What you can judge it on |
|---|---|---|
| Month 1 | Baseline measurement on a frozen prompt set, technical access fixed per named agent, schema and entity foundations, first assets | A baseline you can hold us to. Technical findings closed, verifiable by you. |
| Month 2 | Content built against the questions where competitors are cited and you are not; internal structure; second measurement run | Movement on specific prompts, with run counts and the spread |
| Month 3 | Corroboration — getting named in the third-party sources models already trust for your category. Third measurement. | Citation share against the named competitor set, on the same protocol as the baseline |
| Month 4+ | Prompt-set expansion, defending positions already won, new assets as the buyer questions shift | Trend across runs rather than a single reading |
The honest shape of it: months one and two are largely build, month three is when the measurement starts being worth looking at, and the compounding part — entity strength and third-party corroboration — is what you are paying for from month four onward.
llms.txt, the prompt set, the measurement history. None of it sits behind a platform login you lose access to.If your site fails at the retrieval gate — blocked agents, pages that need JavaScript to render, no schema — a retainer is premature. Those are fixed-scope problems with an end state, and the two-week Foundation Sprint at £1,000 closes them. Buy the retainer when the mechanics work and the problem has become competitive rather than technical.
| Foundation Sprint | Retainer | |
|---|---|---|
| Shape | £1,000 one-off, two weeks | £2,000 – £7,750 a month |
| Fixes | Technical access, schema, top five pages | Everything above, continuously |
| Measurement | Baseline only | Repeated runs, tracked over time |
| Content | Restructure of what exists | New assets every month |
| Entity work | On-site only | Third-party corroboration |
| Best when | The mechanics are broken | The mechanics work and competitors are winning the answer |
Measurement is the unglamorous majority of it. Every prompt is run more than once because assistant output is non-deterministic, and a number from a single run cannot be distinguished from noise. More prompts and more engines means more runs, which is the real reason tiers differ rather than any difference in effort or attention. The protocol is published, run counts included.
Published market retainers run roughly $1,500 to $5,000 a month at entry and $4,000 to $12,000 for active mid-market work. Our monthly tiers are £2,000, £4,250 and £7,750, published in full.
Three months, then monthly with thirty days notice. The three months is not a lock-in device — citation share does not shift meaningfully inside eight weeks, so judging the work earlier than that measures noise rather than progress.
You keep all of it: schema, pages, llms.txt, the prompt set and the full measurement history. Nothing sits behind a platform login you lose. What stops is the measurement runs and new asset production.
Start with the sprint if your site fails at the retrieval gate — blocked agents, render problems, missing schema. Those are fixed-scope problems with an end state. Move to a retainer once the mechanics work and the problem is competitive rather than technical.
Almost entirely prompt count, engine coverage and assets per month. Each prompt is run multiple times on each engine, so measurement cost scales with both. The top tier also includes third-party editorial placement, which is slow and manual.