How do you get budget approved for AI search visibility?
Budget for AI search visibility gets approved when it is framed as revenue already at risk rather than as a new channel to try. Four numbers do most of the work: the share of your pipeline that starts in search, what an AI answer does to the clicks on those searches, the evidence that your buyers are asking engines first, and the specific list of competitors being named instead of you. Pew Research Center found that Google users clicked a traditional result in 8% of searches where an AI summary appeared, against 15% where none did. A March 2026 G2 survey of 1,076 B2B software buyers found 51% now start their software research with an AI chatbot more often than with Google. Where the money comes from is a separate question, and the published answer is that most teams fund it inside the SEO budget rather than as a new channel. No independent benchmark says what share of a search budget should move, and the only survey average on the table is the 12% of digital budget Conductor found at large enterprises. The competitor list is the part you generate yourself, and it is usually what ends the argument.
How do you get budget approved for AI search visibility?
Budget for AI search visibility gets approved when the request is framed as protecting revenue that is already exposed, and refused when it is framed as trying a new channel. Finance teams fund defence of an existing line more readily than they fund experiments, and AI search is genuinely a defence story: the demand was already there, it is being answered somewhere else, and the question is whether your company is in the answer.
The structure that works is short. State how much of the pipeline begins with somebody searching. Show what an AI-generated answer does to the clicks on that kind of search. Show that your buyers specifically are asking engines before they ask vendors. Then show the answers themselves, with the competitors who are named and your own absence. Four steps, two of which you can take from published research and two of which you have to generate from your own data and your own category. On the third step, cite the range rather than the best number: the published surveys of buyer adoption disagree with each other by more than a factor of two, and a finance reviewer who checks will find that out.
What sinks these requests is usually tone rather than content. A paper that opens with the phrase "the AI revolution" is read as a trend piece. A paper that opens with the percentage of last year's closed-won deals whose first touch was organic search is read as a business case, even when the rest of the argument is identical. Lead with your own numbers and keep the industry statistics as support.
One more decision belongs in the paper rather than in the meeting: who does the work. Buying a tracking subscription, hiring, and retaining an outside team are three different cost shapes, and leaving the choice open invites a discussion that has nothing to do with whether the problem is real. Our comparison of agency, tool and in-house models sets out what each one actually covers.
What numbers does a CFO actually want to see?
A CFO wants four numbers, and wants each one to have a source attached. Two of them come from published studies and describe the market. Two of them come from your own systems and describe you. Presenting only the first pair produces a debate about whether the studies apply; presenting only the second pair produces a debate about whether the effect is real. Together they close both exits.
| Number | Where it comes from | What it establishes | Its weakness |
|---|---|---|---|
| Pipeline share that starts in search | Your CRM and analytics, last 12 months | The size of the exposure in your own revenue | Attribution is imperfect and everyone in the room knows it |
| Click loss when an AI answer appears | Pew Research Center, July 2025 | That the exposure is being realised, not forecast | Consumer browsing data, not a B2B sample |
| Share of buyers starting in an AI chatbot | G2 buyer survey, March 2026 | That your buyers are in the new channel already | Software buyers specifically, not every B2B category |
| Competitors named when you are not | Ten prompts you run yourself, this week | That it is happening to your company by name | A small sample, and it varies between runs |
The click-loss figure is the one people underestimate. Pew Research Center analysed one month of browsing data from 900 US adults covering 68,879 Google searches and found users clicked a traditional search result in 8% of searches where an AI summary appeared, against 15% of searches without one, with clicks on links inside the summary itself running at 1% of visits.
The buyer-behaviour figure is the one that converts sceptics. A G2 survey of 1,076 B2B software buyers and decision makers, run in March 2026, found that 51% now start their software research with an AI chatbot more often than with Google, that 71% rely on an AI chatbot somewhere in the research process, up from roughly 60% seven months earlier, and that 69% said an AI chatbot led them to select a different software vendor than they had initially planned.
How do you estimate the revenue at risk from AI search?
Revenue at risk is estimated by working down from pipeline rather than up from traffic, because a traffic number invites an argument about traffic and a pipeline number does not. Start with closed-won revenue over the last year. Take the share whose first recorded touch was organic search. That figure, not your total revenue, is the exposed base, and it is usually large enough that nobody needs it inflated.
Apply a loss rate to the exposed base rather than predicting one. The Pew figures above give a defensible ceiling for informational queries, and a conservative version of the same arithmetic uses only the share of your search-sourced pipeline that begins with a question rather than a brand term, since branded searches behave differently. Present the conservative version. A finance director who finds the pessimistic case already built into the paper stops looking for the catch.
Offset the risk with what the channel is worth when it works, and use somebody else's estimate for it. Semrush's AI search traffic study, published in June 2025 across more than 500 high-value topics, estimated that a visitor arriving from an AI search experience is worth about 4.4 times the average traditional organic visitor, on the reasoning that the engine has already handled the early research and the person clicking through is further along.
Resist modelling a revenue return. The honest position is that you can size what is exposed and you cannot forecast what you will recapture, because citation frequency responds to engine-side changes as well as to your pages. Our page on whether AI search actually sends traffic goes through why volume is the wrong headline metric and what to use instead.
What are other companies spending on AI search visibility?
Other companies are already spending a visible share of digital budget on this, which is useful in a room where being early is treated as a risk. Conductor surveyed more than 250 enterprise executives and digital decision makers for its 2026 State of AEO/GEO report and found they had allocated an average of 12% of digital budgets to answer engine optimisation in 2025, with 94% planning to increase that allocation in 2026 and 97% reporting a positive impact from the previous year's work.
Read that benchmark with its sample in mind. Enterprise digital budgets are not a guide to what a mid-market company should spend in absolute terms, and self-reported satisfaction figures in a vendor-run survey are soft evidence. What the numbers do support is the narrow claim that matters in a budget meeting: this is a line item at comparable companies, not a speculative purchase.
The competitive framing is stronger than the benchmark anyway. Conductor found 32% of the digital leaders it surveyed named answer engine optimisation their single top priority for 2026, more than any other option, which means the companies you are being compared against in AI answers are working on the same problem this year. Position is relative, so standing still is a decline.
Avoid quoting a market growth forecast. Forecasts age badly and invite the reader to argue with the forecast instead of with your case, and there is enough measured behaviour available that you do not need a projection to make the point.
How do you adapt an existing SEO budget for AI search?
Most teams adapt an SEO budget for AI search by funding the work inside the SEO line rather than opening a separate channel, and the published distribution is lopsided enough to settle the argument quickly. In the SEOFOMO State of AI Search Optimization Report 2026, which surveyed 171 search professionals from 36 countries between 17 and 27 August 2026, 54% said their clients or managers had allocated budget "but not as an independent channel" and instead as part of SEO. Only 14% had it as an independent channel, split between 9% where it is larger than the SEO budget and 5% where it is smaller.
| How the budget is structured | Share of respondents | What it implies for your request |
|---|---|---|
| Allocated, inside the SEO budget rather than as its own channel | 54% | The default, and the version least likely to be challenged |
| Not yet, and no plans to | 16% | Still a sixth of the market, so "everyone is doing this" overstates it |
| Not yet, but planned | 11% | The group you are competing with for first-mover position in your category |
| Its own channel, funded above the SEO budget | 9% | Usually a second year, after a pilot produced numbers |
| Its own channel, funded below the SEO budget | 5% | A carve-out, which makes the measurement line visible to finance |
| Do not know | 4% | More common than it should be, and worth checking before you ask |
Ownership follows the money in the same survey. The SEO team or SEO specialists lead the work for 77% of respondents, against 4% with a dedicated AI search team and 4% where nobody is in charge, and the same report records that standalone AI search budgets grew from around 5% to 15% year on year. Read the sample before you quote any of it in a board paper: 171 self-selecting search practitioners are not a random sample of companies, they skew towards agencies and consultants, and the figures describe how the discipline is funded, not how much it should cost.
The practical shape of a reallocation is a carve-out with one new line, not a wholesale shift. Take the existing search budget, move the share that currently buys work with the weakest evidence behind it, and add the one thing the SEO budget has never contained, which is measurement across engines. The same survey names the top challenges as measuring AI search visibility reliably, understanding what drives it, and connecting visibility to traffic, conversions and revenue, which is a strong hint about where the marginal pound belongs.
What share of an SEO budget should move to AI search?
No independent benchmark exists for what share of an SEO budget should move to AI search, so every split you will read is somebody's judgement rather than a measurement. The published numbers come mostly from firms that sell the work, they disagree with each other, and the most widely repeated one is attributed to a report that does not contain it.
The one survey average worth quoting is Conductor's. Its 2026 State of AEO/GEO report, drawn from more than 250 digital leaders at United States enterprises of 500 employees or more, states that enterprises allocated an average of 12% of digital budgets to answer engine optimisation in 2025, and that 56% of those leaders made what the report calls a significant or high investment in it that year (read at source, 9 November 2026). Size-graded versions of that figure circulate widely, usually as 10% of search budget for a small business rising to 30% for a large one, and are credited to the same report. The report as published gives one average and no breakdown by company size, so a size band quoted against it has come from somewhere else. If a proposal in front of you carries one, ask where.
The more useful way to set the number is to price a scope and let the share fall out of it. A prompt set, a list of engines, a run frequency and somebody to read the answers have a cost that can be worked out; a percentage picked first has to be defended afterwards. The table is the order we would move money in, and the reasoning behind each line sits elsewhere on this site rather than in a figure we cannot source.
| Line in the existing budget | What to do with it | Why |
|---|---|---|
| Crawler access, indexing and server-rendered page copy | Protect it before anything moves | Nothing further happens on a page an engine cannot fetch or lift a clean answer out of, which is the first of the five things AI engines test when they choose sources |
| Volume explainer posts on questions the engines now answer themselves | The first candidate to cut | This is the content most exposed to an answer the reader never clicks through from, and the easiest spend to stop |
| Rank tracking as the only reporting line | Keep it, stop treating it as the answer | A position can hold while the answer above it names somebody else, which is the situation most teams arrive with |
| Cross-engine citation measurement | The new line, and usually the only genuinely new cost | The SEOFOMO survey of 171 search professionals named measuring AI search visibility reliably as the leading challenge in this work, and it is the one thing an SEO budget has never contained |
| Original data, documentation and named customer work | Increase if anything is left | Corroboration elsewhere on the web is part of what decides whether a claim on your page survives into an answer |
Two tests will tell you whether a proposed split is serious. Ask what it buys that the current budget does not buy today, and ask what it stops. A reallocation with no answer to the second question is an increase by another name, and a finance team will work that out faster than you will.
What should a reallocated AI search budget actually buy?
The survey is specific about what the money should buy, and none of it is exotic. Respondents credited expanding content to cover more of the questions their category gets asked, improving crawlability and accessibility, and restructuring content so answers are easier to lift, which is largely the same work an SEO team already knows how to do. That is the honest case for funding it inside SEO, and it is also why a request framed as a brand new discipline invites more scrutiny than one framed as a redirection of existing effort.
Set the expectation about attribution in the same paper, because it is the part that goes wrong at the next review. In that survey 46% put AI search at 0% to 5% of site revenue and 35% said they had no idea what it contributes, while the share attributing more than 5% of revenue to it rose from 8% to 19% in a year. Promising a revenue line in quarter one against those numbers is how a reallocation gets reversed, and our guide on telling whether your AEO is working sets out what to commit to instead.
What should a procurement team ask for when buying AI visibility tracking across markets?
A procurement team should ask for the engine list, the market list and the run schedule before it asks for a price. Those are what set the price, and they are the parts a proposal most often leaves vague. A quote for "AI visibility tracking" with no engine list is not comparable with anything, and two vendors can differ by an order of magnitude on the same brief without either being wrong.
Four things drive what multi-market tracking costs, and all four are scope rather than rate. The number of markets, because each one needs its own prompt list in its own language. The number of engines per market, which is not uniform: a Chinese, Russian or Korean market is answered by a domestic engine that most platforms do not query, as our page on whether AI search recommends different vendors in different countries works through. The number of prompts, and how many times each is repeated, since a single run per market is an anecdote rather than a measurement. And whether somebody reads the answers, because a dashboard that nobody interprets costs less and tells you less.
Ask for the arithmetic in the proposal rather than a headline figure. Four markets at 20 prompts across three engines, run four times a month, is 960 answers a month, and a vendor who cannot tell you that number has not scoped your brief. Ask also what happens in a market where no tool reaches the engine. A vendor with a workable answer will describe a local arrangement running the same prompt list; a vendor without one will change the subject.
Two clauses are worth insisting on. That the prompt set is yours and leaves with you, in full, with the run log, because a programme whose history is locked inside a vendor dashboard cannot be re-tendered. And that the method is held still for the length of the contract, since changing the prompt set or the counting rule mid-programme produces a movement that reflects neither the market nor the work. Our comparison of agency, tool and in-house models sets out what each arrangement actually covers, and SIGNALS figures are on the pricing page rather than in articles.
How do you check whether an AI visibility vendor is legitimate?
Check whether an AI visibility vendor is legitimate by making it show you its method and its data exports, not by hunting for reviews. Reviews are a weak signal in a category this young: an empty profile often means nothing more than that the firm is newer than the engines it measures, and a full one is frequently a software listing for a product that is only part of what you would be buying. The checks that separate a real supplier from a reseller of somebody else's dashboard are all things a vendor can demonstrate in a single call.
Start with what the vendor will put in writing before you sign. Mike King's list is the most usable version of this and it is short: which engines the thing covers, how many prompts it samples and how often it collects data; the path from a specific finding to a proposed action, with the evidence behind it; the review your own team has to do before any automated fix goes live; and what you can export, how much history you keep and what happens when the subscription ends (Mike King, AI Search Agency vs Platform, 4 November 2026, read at source 9 November 2026). He runs an agency and argues in that piece that a platform alone is not enough, which is worth knowing while you read him. The questions survive the conflict of interest because a vendor who cannot answer them has not scoped your brief.
Three further checks are worth the ten minutes they take. Ask for a prompt set and a run log from real work, redacted if it belongs to somebody else, because a firm that has never run a baseline cannot produce either. Ask who does the page work, since a diagnosis with no delivery attached leaves you with a report and the original problem. And ask what the vendor will not claim: anyone promising a number of citations by a date is selling something the engines do not let them control, which is the single most reliable tell in this market.
What a missing rate card means is less than buyers assume. Plenty of credible firms in this lane scope by engagement and publish nothing, and plenty of thin ones publish a tidy table. Read the absence as a reason to ask what drives the number rather than as a mark against the firm, and our page on whether you need to publish pricing to get recommended by AI covers the same question from the vendor's side. SIGNALS is one of the firms in this category and publishes this page, so treat the paragraphs above as our account of what a buyer should demand, including from us. Our figures are on the pricing page rather than in articles.
How do you answer "isn't this just SEO with a new name"?
Answer the SEO objection with the measurement rather than with a definition, because the definitional argument is unwinnable and the empirical one is not. The useful reply is that the two produce different winners on the same query, which is checkable in ten minutes in front of whoever asked.
The published version of the same point is blunt. The ConvertMate GEO Benchmark 2026, an observational study across 8,000 domains, found that 83% of AI citations come from pages outside Google's top 10 results. A company can hold the top ranking for a term and not be in the answer that now sits above it, which is the situation that brings most people to this problem in the first place.
There is a second, more uncomfortable version of the objection: that the current SEO supplier should simply do this. Sometimes they can, and the honest test is whether they measure citation share across engines today and can show you the log. Our guide to choosing an AEO agency, or an SEO agency that can do AEO covers the questions that separate the two.
Where the objection is fair is on the work itself. Much of what improves AI citation is unglamorous page work that a competent content team could do, given a diagnosis of which pages and what to change. Conceding that openly makes the rest of the case more credible, and it points the budget at the part nobody currently owns, which is measurement.
What should you ask for in the first budget cycle?
In the first budget cycle, ask for a 90-day pilot with a measurement commitment attached, rather than an annual programme. A pilot is approvable at a level of authority that does not require a committee, it produces evidence rather than opinions, and it converts the question from "is AI search real" into "did the number move", which is the question you want to be answering next quarter.
Scope the pilot around a baseline rather than around output. The deliverable is a fixed prompt set run across ChatGPT, Claude, Perplexity and Google AI Overviews, measured twice before anything changes so the variance is known, then a prioritised set of page fixes and a re-measurement. Saying up front how much two identical baseline runs are expected to differ protects you later, when they do, and our page on knowing whether your AEO is working sets out the thresholds.
Name the metric you will be judged on before somebody else picks one for you. Citation share on a fixed prompt set is the honest choice: it responds to the work, it can be compared month to month, and it does not collapse when referral traffic stays small. Referral sessions and AI-sourced leads belong in the report as secondary lines, clearly labelled as lagging.
Bring evidence to the meeting rather than a proposal to run one. Ten prompts through four engines, with the answers pasted in, takes an afternoon and is the single most persuasive attachment you can send. If you would rather have the baseline run properly, a free visibility assessment produces the same artefact across your real buyer questions and reports who is cited, from which page, and where you are being passed over.
What else do people ask about funding AI search visibility?
How much should a B2B company budget for AI search visibility?
Enough to run a real prompt set across four engines every month and to change pages on the basis of what it says. Conductor's 2026 survey of more than 250 enterprise executives found companies had allocated an average of 12% of their digital budgets to answer engine optimisation in 2025, which is a useful benchmark for a large organisation and a poor one for a 20-person company. Scope, the number of product lines and how much of the page work you outsource are what actually move the number. SIGNALS figures are on the pricing page rather than in articles.
Should the budget come out of SEO or be new money?
Usually some of each, and in practice most teams put it inside SEO: 54% of the 171 respondents to the SEOFOMO 2026 survey had budget allocated as part of SEO rather than as an independent channel. Taking it all from SEO invites the argument that the work is a rebrand of what somebody is already paid to do. Asking for all of it as new money invites the argument that it is unproven. A first phase funded largely from existing search budget, with a modest increment for the measurement that does not exist today, tends to get through faster than either extreme.
What is the strongest single piece of evidence to put in front of leadership?
A screenshot of an AI answer to a real buyer question in your category, with three competitors named and your company absent. Numbers from published studies establish that the channel exists. The screenshot establishes that it is happening to you, and it is the item people remember when the meeting is over. Run ten questions before writing anything and you will usually find at least one.
How do you avoid promising a return you cannot deliver?
State the mechanism and the measurement separately from the outcome. What you can commit to is a baseline, a fixed prompt set, a monthly reading of citation share against named competitors, and a list of page changes made. What nobody can commit to is a number of citations by a date, because citation frequency moves on engine-side changes as well as on your work. Writing that distinction into the proposal yourself is more persuasive than having a finance director find it.
What if leadership says it is too early?
Answer with the adoption data rather than with a prediction. Forrester's State of Business Buying 2026, a survey of nearly 18,000 global B2B buyers, found 94% now use AI in their buying process, and the same G2 survey that put 51% of software buyers starting in an AI chatbot recorded that share rising from 29% in early 2025. Early would have been 2024. The current question is whether a competitor locks in the position first, and a 90-day pilot is a cheaper way to settle it than a strategy debate.