Decision°

What does generative engine optimisation cost?

· 8 min read ·
What does generative engine optimisation cost?

Across the market in 2026: AI visibility tooling runs roughly £200 to £2,000 a month, freelance GEO £2,000 to £6,000, agency retainers £4,000 to £15,000, senior advisory £8,000 to £35,000. An in-house hire is £60,000 to £95,000.

The ranges, and what sits inside each

These are observed market ranges for the UK and US in 2026, not our rates. They are wide because the label covers very different scopes, and the first job of any pricing conversation is establishing which scope is being quoted.

Market ranges, monthly unless stated
OptionTypical rangeWhat you actually get
AI visibility tooling only£200 – £2,000Monitoring across assistants, refreshed regularly. No judgement, no work done. Priced on prompt volume and competitors tracked.
Freelance GEO specialist£2,000 – £6,000One person, part-time, usually content restructuring and technical fixes. Rarely covers off-site citation sources.
Agency retainer£4,000 – £15,000A team, usually delivery-led. Scope varies enormously; this is where the schema-and-content-only offers cluster.
Senior advisory£8,000 – £35,000Strategy and direction with your team executing. Priced on seniority and breadth rather than output volume.
In-house hire£60k – £95k salaryPlus tooling, plus the ramp. Realistically six months before the role pays back.
One-off audit£1,500 – £8,000A baseline and a roadmap. Some firms, including us, do a version of this free.

What drives the number up

  • How contested your category is

    A category with entrenched review-platform incumbents and a busy subreddit is far more expensive to move than a quiet niche. This is the single biggest variable and it is rarely priced explicitly.

  • Whether off-site work is in scope

    The cheapest offers are on-site only. They are cheaper because they address the minority of citations. Off-site programmes need different people and run over quarters.

  • How many platforms and markets

    Measurement cost scales with prompts times runs times platforms times regions. Tracking five assistants in three markets is not three times the cost of one, but it is not one either.

  • Seniority

    The judgement calls in this category, which prompts matter, whether an answer is materially wrong, where a programme should stop, are what you are actually buying at the top of the range.

  • Whether your foundation is sound

    A site with weak organic search, no schema and blocked crawlers needs remedial work before any AI-specific programme can produce anything.

What you should not be paying for

  • llms.txt implementation

    A May 2026 log analysis of 137,000 domains found 97% of llms.txt files received zero requests, and Google stated in the same month that the file has no effect on Search visibility. It is one of the most commonly sold GEO line items and the evidence says it does nothing.

  • Guaranteed citations or placements

    Model outputs are non-deterministic and change between versions without notice. A guarantee here is describing something that cannot exist, and its presence in a proposal tells you something about the rest of it.

  • A blended AI visibility score as the deliverable

    One number averaging several platforms hides which platform is failing, which is usually the thing you needed to know. Ask for per-platform reporting.

  • Review generation as a volume deliverable

    Buying or incentivising reviews beyond platform rules violates every marketplace's terms, is detectable, and the exposure sits with your brand rather than your supplier's.

  • Bing SEO sold as Copilot optimisation

    Copilot's web answers are grounded in the Bing index and there is no separate Copilot crawler. The work is legitimate; the markup for calling it AI is not.

How to sequence spend on a small budget

  1. 1Spend nothing and check your crawler directives. Blocking a retrieval crawler removes you from answers entirely, and it is an afternoon to verify.
  2. 2Spend nothing and run the manual visibility check: ten questions, five platforms, three runs. About an hour, and it tells you which problem you have.
  3. 3If you are absent everywhere, the problem is off-site, and tooling will not fix it. Prioritise review-platform hygiene, which is cheap, before anything else.
  4. 4If you are named but never cited, the problem is extractability and third-party sources describing you. That is content work you may already be paying for.
  5. 5Only then buy monitoring, and only at the point where doing it by hand has become a job rather than a task.
  6. 6Bring in senior help for the judgement calls, not for the execution, unless you genuinely have no team.

Our own pricing

Surge45 prices on request, tiered by stage, because the scope varies enough that a published rate would mislead more often than it helped. What we will say publicly is that the free AI Visibility Audit is genuinely free, has no obligation attached, and you keep the prompt set afterwards whether or not we work together.

We have published this page with market ranges rather than our own because the question a buyer is asking is what this costs in general, and answering that honestly is more useful than a rate card.

Our position

If a supplier will not tell you which of the six scopes in the first table they are quoting, that is the conversation to have before the price.

Related questions

Start with the free audit and price the work after

What does GEO cost? Market rates for AI search work | Surge45