can we buy the strategy on its own and implement it ourselves?

Not as a separate product, but you already get the first version of it for nothing: the audit ends with a dated 90-day plan in priority order, and plenty of companies take that plan in-house. What the retainer adds is the frozen panel re-run every two weeks, the revision of the plan when the numbers say it was wrong, and the implementation. If your team can do all three, the audit plan is the strategy and you should use it.

how do you decide which prompts to commit to?

From the audit's logged answers. For each category prompt we know which competitors were named, from which sources, on which engines. We commit to the prompts where the cited sources are ones you can realistically enter and the competitor set is thin, and we list the prompts we are not committing to with the reason. The commitment is scoped against your measured baseline, never sold before it exists.

what if our competitors in the answers are not the ones we compete with?

That is the normal case, and it is one of the most useful things the strategy tells you. The models name whoever is verifiable and present in the sources they read, which is often not the company that ranks or the company you lose deals to. The plan targets the answer's competitor set, because that is the list your buyer sees, while the classic SEO layer inside the retainer keeps working the ranking one.

GEO strategy: the prompts worth winning

GEO strategy is the stage between measuring and building. The audit says what six engines answer about you today and who they name instead; the strategy decides which of those answers are worth taking, in what order, and by what route. It is a short document with a long shelf life: a frozen prompt panel, a competitor map built from the answers rather than the rankings, a list of the sources the engines actually cite in your category, and a sequenced plan that respects the one rule the dataset keeps teaching, readable first, verifiable second, quotable third.

It is not sold on its own, and we would not know how to. There is no useful way to scope a strategy before the audit has measured where you stand, which is why stage two only exists inside the retainer. What you get before you pay anything is the 90-day plan at the end of the audit: the first draft of the strategy, complimentary, and yours to keep.

why the strategy is not a keyword list

The failure the strategy exists to prevent is publishing more of what already was not working. In the Manimama case the firm arrived with 1,004 blog posts, none in Google US top-3, while the competitor ahead of it had 6.8 times less content (baseline 22 June 2026, the case). The engines' complaint was specific: anonymous articles, no visible lawyers, nothing verifiable. The strategy that came out of the analysis was one sentence, out-verify them rather than out-publish them, and it is the sentence that moved the panel from 0 of 20 prompts to 20 of 20 across four engines by August 2026.

That is what stage two is for. Across the 100 audits in our published dataset (April to July 2026), 93% of brands were known to the engines by name and 4.5% were named when a buyer asked who to pick. Closing that gap is rarely a content-volume problem, and the strategy is where we say so in writing, with your numbers, before anyone writes a page.

what you get

the strategy, section by section
no.	section	what is inside
01	the prompt panel	Branded probes and category prompts, worded once and frozen, per market and language. This is the panel every later review re-runs, so its wording is the most consequential decision in the engagement.
02	the competitor map	Two lists, side by side: who ranks for the keyword, and who the models name in the answer. They are frequently different companies, and the second list is the one we plan against.
03	the source map	Every source the engines cited in your category answers, logged from the panel: directories, roundups, comparison pages, communities. Which you can realistically enter, and which you cannot.
04	the entity plan	What a model needs to confirm you are one specific company: the records to create, the facts to reconcile across profiles, the name collisions to resolve.
05	the page map	The page types the engines quote for your money queries, which of them you have, and the order to ship the rest in.
06	the sequence and the targets	Sprint by sprint, with exit criteria, and the target prompts we commit to. The guarantee lives here, scoped against the measured baseline.

how we build it

Read the panel, not the rankings. Every category answer from the audit is already logged with its sources. The strategy starts by counting: which competitors appear most, from which sources, in which positions, on which engines.
Find the bottleneck layer. Readable, verifiable or quotable. The band your health score sits in usually says which: below 25 it is engineering, 26 to 40 it is entity work, above 40 it is other people's pages (how to read a health score).
Choose the prompts to commit to. Not every prompt is winnable in 90 days. We pick the ones where the cited sources are enterable and the competitor set is thin, and we say which ones we are not committing to and why.
Sequence the work. Technical fixes first because everything else is invisible until they land, entity records second because citations do not hold on an unverifiable brand, then content and authority in parallel.
Set the exit criteria. Each sprint ends with something a developer can verify and a re-run of the panel. If a sprint has no measurable exit, it is not in the plan.

how it is measured

The strategy is judged on the panel it defined. Every two weeks the same prompts are re-run on the same engines and reported as a delta against the logged baseline: brand recall, category recall, position in the answer, and which source the engine leaned on. Because the panel is frozen, two dates are comparable, and because it is a series, a single lucky run does not count. The Manimama panel is a public example of the format: 800 answers per pass, dated run by dated run, including the rollback.

When the numbers do not move, the strategy is the document we go back to. Usually the layer that failed is citability, because it depends on third parties publishing; sometimes the prompts were wrong. Either way it is revised in writing, and two consecutive flat reviews are a reason to stop rather than to re-sell.

what it costs

Stage two is delivered inside the full-cycle retainer, which starts from $3,000 per month (September 2026); the final figure is set after the audit. It is not a separate line item, because it is not a separable piece of work: a strategy without the measurement that precedes it and the sprints that follow it is a slide deck. The audit, which produces the first version of the plan, is complimentary. What moves the retainer above the floor is on the pricing page: markets and languages, brands and domains, who implements, and the state you start from.

where it sits in the loop

Stage one is the audit. Stage three, the first sprint the strategy schedules, is almost always technical GEO. Stages four and five, content and SXO and links and authority, are what the page map and the source map turn into. The whole loop is on the services page.

GEO strategy