How much does AI search optimisation cost? A buyer’s guide to scope, price and red flags.

The short answer.

AI search optimisation costs whatever its scope costs: the number of buyer questions and AI engines covered, the content and off-site work included, the technical fixes and the measurement. The largest published dataset we found, an October 2026 study by the agency First Page Sage, puts the median monthly retainer at $6,850, but quotes in that same data run from under $2,000 to $60,000 and more a month.

So the useful question is not “what does it cost?” but “what exactly is in this proposal, and how will we both know it worked?” A cheap proposal with no baseline is expensive. A large one that is mostly reporting is too. The rest of this guide gives you the tools to tell them apart.

What you are actually paying for.

Most of the work called AI search optimisation, GEO or AEO is not a new discipline with secret levers. Google says so plainly in its own guidance on AI features: there are “no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary”, and “no special schema.org structured data that you need to add.” A page has to be indexed and eligible to appear in normal search. The same guidance says these features may use “query fan-out”, running several related searches across subtopics to find supporting pages.

That tells you where the money really goes. You are paying for three things: clear pages that answer the questions your buyers ask, a reputation on other websites that engines read when they compare providers, and someone to measure whether you are being named. If you want the vocabulary first, our explainer on AEO, GEO and SEO covers the differences in plain terms.

It also tells you what you should not pay a premium for. A proposal that prices “AI-ready schema” or an “LLM file” as a headline deliverable is selling something the largest search engine says you do not need. Structured data that matches your visible text is good practice. It is not a strategy.

The five things that drive cost.

When we compare proposals, the price difference almost always traces back to five variables. Ask each agency to state its assumption for every one. If they cannot, they have not scoped the work; they have picked a number.

Questions and engines set the size of everything else

Ten buyer questions in one engine is a small job. Fifty questions across ChatGPT, Perplexity, Google AI Overviews and Gemini, in two countries, re-run monthly, is a different job. Engines draw on different sources, so each extra engine adds research, not just a row in a report. Start by writing down the twenty questions your buyers actually ask before they shortlist you. That list is the most useful thing you can give any agency, and it makes their quotes comparable.

Content and off-site work are where most hours go

Writing a page that answers a buyer’s question with dated evidence takes real time, as does getting accurate profiles and genuine reviews on the sites engines read. If a proposal is cheap and also promises a lot of both, ask who is doing the work and how long each piece takes. Volume produced with AI tools and no added value is not a bargain: Google’s spam policies name “using generative AI tools or other similar tools to generate many pages without adding value for users” as scaled content abuse.

What the published market data says.

There is no independent, audited survey of AI search pricing. What exists is published by agencies, so read it as a rough map, not a benchmark. The most detailed we found is First Page Sage’s GEO and AEO pricing report, published 7 October 2026. It compiled 1,146 price quotes from 214 agencies between 6 July and 2 October 2026, drawn from pricing pages, proposals and signed contracts. First Page Sage sells these services itself and includes its own pricing in the data.

A few details in that report matter more than the headline median:

• Industry shifts the price. The report gives B2B SaaS a median of $8,640 a month, with a typical range of $4,500 to $20,000.
• Published prices understate the market. Only 23% of the 214 agencies published any pricing, and published pages lean towards entry tiers.
• Performance pricing is rare, at 5% of quotes, and the report notes buyers struggled to agree how AI mentions would be counted.

For a second view, WebFX, a US agency, publishes ranges on its GEO pricing page (published 15 October 2026): $1,500 to $5,000 a month for small businesses, $5,000 to $25,000+ for mid-sized ones, $25,000 to $50,000+ for enterprises, and $5,000 to $50,000 for one-off projects. The page states no method behind the ranges. The two sources broadly agree on shape and disagree on detail, which is what you would expect in a market this young. Use the ranges to sense-check a quote, never to choose one.

Audit, project or retainer: which shape fits.

Most proposals come in one of three shapes. Each has a job it does well and a way it goes wrong. In First Page Sage’s data, one-time audits had a median of $4,300 and usually ran two to six weeks; projects or sprints a median of $14,500 over 30 to 90 days; and custom retainers a median of $7,200 a month, most often on six-month terms.

For most $1M to $50M B2B companies, the sensible order is the same: a baseline first, then a short project on the gaps the baseline found, and a retainer only if the project shows the work moves something you can measure. If an agency will only sell you the retainer, ask why.

What a good scope document contains.

A scope document is the thing you will hold the agency to in month four, when memories differ. It should be specific enough that a stranger could check whether the work was done. These are the sections we expect to see, and the ones we put in our own scopes:

1. The baseline. The buyer questions tested, the engines, the dates, the country, and who was named, cited or missing. Kept as screenshots or exports you can open.
2. The question set. The exact questions that will be re-tested, agreed with you, and how often.
3. Named deliverables. Pages to write or rewrite by URL or title, off-site profiles by site name, technical fixes by issue.
4. Who does what. Which changes the agency makes, which need your developer, and what access they need.
5. Measurement. What counts as progress (being named, being cited, referral visits, enquiries), the tool, and how you get the raw data.
6. Fact sources. Where claims about your company come from, and who approves anything that names a client or a number.
7. Exit terms. Notice period, what you keep, and how profiles and accounts created in your name are handed back.

Item seven is the one buyers forget. Review profiles, directory listings and tracking accounts created during the work should be in your company’s name, with your login, from day one.

Red flags in a proposal.

Google’s own advice on hiring an SEO is blunt: nobody can promise a top ranking in its results, and if someone promises first place, “find someone else.” The same logic applies with more force to AI answers, which change from one run to the next. A study by SparkToro and Gumshoe.ai, reported by Search Engine Journal on 30 January 2026, ran 12 brand-recommendation prompts 2,961 times across ChatGPT, Claude and Google’s AI. ChatGPT and Google returned the same list of brands in under 1% of repeated runs, and the same list in the same order in under 0.1%.

None of these flags means the agency is dishonest. Some are simply new to the work and borrowing language from old SEO sales decks. But each one tells you the proposal has not yet been written around evidence you can check.

How to compare proposals fairly.

Proposals are written to be hard to compare. One quotes a monthly fee with a long list of activities, another a project price with a short list of outcomes. Put them on one sheet before you look at the totals.

Normalise the scope first

For each proposal, write down the number of buyer questions, the engines, how often they are re-tested, the pages produced per month, the named off-site sites, and the hours of technical work. Then divide the fee by the deliverables you actually care about. A proposal that looks twice the price can be cheaper per useful page once you strip out reporting time.

Check the evidence behind their own claims

Ask each agency for one result with its source, its metric name and both date windows. This is how we present ours: for FullTilt Team Development, impressions inside Google’s Generative AI features rose 6.7x, from 3.45K to 23.1K, comparing the last three months with the previous three (Google Search Console, captured 4 September 2026). The FullTilt case study shows the screenshot. If an agency’s result has no source or no dates, treat it as marketing.

Ask how they will connect visibility to enquiries

Being named is not the goal; enquiries are. A good proposal says how AI referral visits will be tracked in analytics and matched to the enquiries that follow. Our guide on turning AI visibility into qualified enquiries covers what to measure once you are being named.

The proposal scorer.

Open a proposal and tick each item it clearly includes. You get a score out of 100 and a list of what is missing, to send back as questions. Score each proposal separately. Nothing you tick is sent anywhere.

The weights favour the baseline, the question set and checkable measurement, because those decide whether you can ever tell if the money worked. A proposal under 60 is not necessarily a bad agency; it is an unfinished scope. Send it back with the missing list.

Spending a small budget well.

If budget is tight, the first spend should be on finding out what the engines already say about you, because the cheapest fixes are usually there. We know because we tested ourselves. On 6 November 2026 we asked ChatGPT, Perplexity and Google’s AI Overview which agencies help B2B companies with AI search visibility in the US. ChatGPT named eight agencies, Perplexity ten and the AI Overview six. We were in none of them, and only one agency appeared in all three.

When we then asked ChatGPT about us by name, it called us “worth shortlisting”, but it quoted an outdated 4.17x figure for a client result that our site now reports as 6.7x, and it mixed up our founding year: 2015 on our own profiles, 2011 on some older directories. Neither problem needed a retainer to find. Both are fixed by making our facts consistent across the sites engines read.

A sensible order for a modest budget

• Run a baseline on twenty real buyer questions, in the engines your buyers use. Our guide on how to check your AI search visibility shows how to do it yourself.
• Correct wrong facts about your company on your own site and the main profiles first.
• Rewrite the two or three pages that should answer your buyers’ most common questions.
• Only then pay for ongoing content or off-site work, with the baseline as the yardstick.

If you want to understand why engines leave a company out before deciding what to buy, our diagnosis of why brands go missing from AI shortlists covers the mechanics, and the AEO and GEO visibility guide goes deeper.

Questions owners ask.

How much does AI search optimisation cost per month?

Published agency data varies widely. First Page Sage’s October 2026 report, built from 1,146 quotes, gives a median retainer of $6,850 a month, with mid-market tiers around $4,500 to $11,000 and enterprise work above $22,000. Your price should follow your scope: questions, engines, content, off-site work and measurement.

Is GEO more expensive than SEO?

In First Page Sage’s 2026 data the median GEO and AEO retainer was higher than the median traditional SEO retainer. In practice much of the work overlaps, because Google says AI features need no special optimisation beyond normal search eligibility. Ask what is genuinely new in the scope and what is ordinary SEO.

Can an agency promise to get us into ChatGPT answers?

No. Nobody controls what an AI engine says, and answers change between runs. A 2026 study reported by Search Engine Journal found ChatGPT repeated the same brand list in under 1% of runs. An honest agency will promise work and measurement, not placement.

Should we start with an audit or a retainer?

Start with a baseline or a short audit. It gives you a dated starting point, shows which gaps are worth paying to close and makes any later proposal easy to compare. Move to a retainer only when a defined project has shown measurable movement.

What should be in the monthly report?

The questions re-tested, the engines, the dates, whether you were named or cited in each answer, what was delivered that month, and AI referral visits and enquiries from analytics. You should be able to open the raw answers behind every number.

Sources: Prolinkage’s own AI engine test, 6 November 2026 (ChatGPT, Perplexity, Google AI Overview). Client result: FullTilt Team Development, Google Search Console Generative AI features impressions, last 3 months vs previous 3 months, captured 4 Sep 2026. External: First Page Sage, GEO and AEO pricing report, 7 Oct 2026; WebFX, GEO pricing, 15 Oct 2026; Google Search Central, AI features and your website; Google Search Central, spam policies; Google Search Central, do you need an SEO?; Search Engine Journal on SparkToro and Gumshoe.ai, 30 Jan 2026; US FTC, fake reviews rule, 14 Aug 2024. All accessed 7 November 2026.