Is AEO worth it for your company, or should you wait?
AEO is worth paying for when three things are true at once: buyers in your category already use AI assistants to build a shortlist, your pages can be fetched and read, and somebody internally can approve a claim within a week. Semrush's survey of 622 US B2B professionals, published 8 July 2026, found 66% of the AI users among them regularly using AI to research products, vendors or solutions, and 62% using it while actively comparing vendors, so the first condition holds in most categories. Where it fails, or where the other two fail, waiting is the better call and this page says so in detail. We sell this work, which is a reason to read the section on when not to buy it closely.
Is AEO worth it for your company, or should you wait?
AEO is worth it for your company if buyers in your category ask AI assistants which vendors to consider, and it is worth waiting if they do not or if your own site cannot be read. Those two tests do most of the work. Everything else is a question of sequencing and who owns it.
Most B2B categories now pass the first test. Semrush surveyed 643 US B2B professionals in March and April 2026, kept 622 valid responses, and reported that among the 519 who use AI for work, 66% regularly use it to research products, vendors or solutions, with another 29% doing so occasionally. The same survey put 72% at the early research stage and 62% at the point of actively comparing vendors.
The second test is where companies fail without knowing it. If your key pages render only in JavaScript, sit behind a form, or are not linked from anywhere, the engines cannot use them and no amount of writing changes that. Fix the mechanical problem and you may find you need less help than you thought.
The third test is organisational and it is the one that quietly kills engagements. This work changes what your website claims, every month, and every change needs somebody who can say yes. A company where legal takes six weeks to clear a sentence should either solve that first or plan for a much slower programme, which we cover in doing AEO when legal and compliance approve every page.
Who should not pay for AI search visibility yet?
Five kinds of company should not pay for AI search visibility yet, and we turn away versions of all five. Saying which ones is more useful than a list of who it suits, because the failure cases are specific and the success case is general.
First, a company with no indexable pages about what it sells. A site of five pages, three of them images, has nothing to score and nothing to quote. Writing those pages is ordinary content work and you do not need an AI visibility firm to brief it. Second, a company whose category has no demand in AI answers at all, which happens with genuinely new product categories where buyers have no name for the thing yet. Our page on whether AEO works if nobody searches your category covers how to test that and what to do instead.
Third, a company whose pipeline is entirely referral or founder-led and is not trying to change that. The work still has value, because referred buyers check you, but it is not the first thing to buy. Fourth, a company in the middle of a replatform or a rebrand. Rewriting pages that are about to be thrown away wastes the quarter, and a migration can cost you visibility on its own, as our page on whether a redesign hurts AI visibility sets out.
Fifth, and most common, a company that cannot name the questions its buyers ask. Not the keywords: the questions, in the buyer's words. If nobody in the company can produce fifteen of them, the right first step is twenty customer calls, not a retainer. A programme built on invented questions optimises for an audience that does not exist.
How do you tell whether your buyers are asking AI about your category?
Find out whether your buyers ask AI about your category by asking them and by running the prompts yourself, in that order. Both take days, not months, and either one can settle the question before you spend anything.
The direct method is a question on your enquiry form and in your discovery calls: how did you come across us, and did you use an AI assistant while looking. Self-reported attribution is unfashionable because it is imprecise, and it is also the only thing that sees a buyer who asked ChatGPT for options, then typed your name into Google. Kevin Indig, writing on 19 October 2026 after running 632 timing tests across four brands, concluded that he could not detect hidden AI traffic in analytics at all without self-reported attribution.
The indirect method is to behave like your buyer. Write out fifteen to twenty questions a prospect would type while choosing a supplier in your category, run them across ChatGPT, Gemini, Claude and Perplexity, and record two things: whether any vendor is named, and whose pages are cited. A category where the engines refuse to name vendors is a category where this work is early. A category where they confidently name three competitors is a category where you are already losing something.
One warning about reading those answers. The same prompt produces different answers on different days and for different users, so one run tells you very little. Our pages on how many prompts you need and why AI answers change every time explain how much repetition a reading needs before it means anything.
What has to be true on your own site before an engagement can work?
Before an engagement can work, your site has to be reachable, readable, and about the questions buyers ask. The order matters, because the first two are prerequisites and the third is the actual work.
Reachable means the engines' crawlers can fetch your pages and are not blocked, and that the pages you care about are linked from somewhere. Readable means the content is in the HTML rather than assembled by a script after load. Both are cheap to check and both are common failures, which is why our readability checklist exists as a free page rather than as part of a paid audit.
The reason these come first rather than later is that retrieval dominates everything downstream. Chen and Koudas, in a study of 158,099 answers and 468,037 citations published 10 October 2026, found the engine's own search step explaining about 32% of the variation in which pages get cited, against 2.8% for the properties of the pages. A page that cannot be retrieved scores zero regardless of how well it is written.
The third requirement is the one that cannot be bought cheaply: a page for each question, kept current. Seer Interactive's study of 7,683 cited pages, published 24 July 2026, found 75% of them updated within the last year. Maintenance is the commitment, which is why this is sold as a retainer rather than a project, a distinction we argue out in is AEO a project or a retainer.
What does the honest upside look like if you do buy?
The honest upside is being named and quoted when a buyer asks an engine who to consider, and it is not a traffic channel. Anyone promising you a traffic line that replaces organic search is describing something nobody has measured.
Take the traffic question first, because it is where expectations go wrong. SE Ranking's study of 101,574 websites, published 18 June 2026, found AI platforms accounting for 0.32% of all website traffic in 2026, up from 0.24% in 2025, with Google still sending about 134 times more visitors. Within that sliver, ChatGPT carried 74.78% of AI referrals, Gemini 11.56% and Perplexity 7.23%. The volume is small and growing fast, and planning around the volume misses the point.
The value is upstream of the click. In the Semrush survey above, only 7% of respondents said they notice a vendor in an AI response because they recognise the name, while 75% fully or mostly trust AI vendor recommendations. Read together, those two numbers describe a channel where being present matters more than being famous, which is unusually good news for a company that is not the category leader.
Our own research says something similar about where the citations land. Across six categories in our study of how four AI engines answer process skid sourcing questions, between 73% and 89% of citations pointed at suppliers' own capability pages rather than at directories, marketplaces or listicles, and no single supplier was named by all four engines in any contestable category. Your own pages are what gets quoted, and nobody owns the answer yet.
Should you wait, do it in-house, buy a tool, or hire a firm?
Wait, do it in-house, buy a tool or hire a firm: the right answer depends on which of the three conditions above you fail and how much internal capacity you have. We publish this page and we sell the fourth option, so read our row the way you would read any supplier's account of itself.
| Option | Who it is for | When not to pick it |
|---|---|---|
| Wait and fix fundamentals | Companies mid replatform, with no indexable content, or unable to name their buyers' questions | When competitors are already named in answers for your category and you are not |
| Do it in-house | Teams with a category-literate writer, a developer with an hour a week, and someone to run prompts monthly | When nobody can protect the time, or when the same person owns demand gen and would deprioritise it |
| Buy a tracking tool | Teams that know what to fix and need to see movement across engines | When you need the pages rewritten, because a tool reports and does not write |
| Hire a firm, including us | Companies with pages worth fixing, buyers who use AI to shortlist, and no internal capacity to do the work monthly | When your category has no AI demand yet, when your site cannot be crawled, or when you want a one-off project and will not maintain it |
Cost is the question this table does not answer, deliberately. What moves a quote is scope, how many pages need rewriting, how much research the category needs and whether the engines are also wrong about your facts; our current figures live on the pricing page rather than inside an article, where they would go stale the day they changed. If you would rather see which of the four rows you are in before deciding, our free visibility assessment scores your pages and names the buyer questions you are absent from.
What else do people ask about whether AEO is worth it?
The five questions below come up in first conversations with companies deciding whether to fund this work now, next quarter, or not at all.
How do I know if AEO is worth it for my company?
Three conditions decide it: buyers in your category use AI assistants while choosing vendors, your pages can be fetched and read, and somebody internally can approve claims every month. Where all three hold, the work pays. Where any one fails, fix that first, because an engagement cannot supply it for you.
Is AEO a waste of money if we get all our business from referrals?
Not a waste, but a weak first priority. A company whose pipeline comes from referrals and outbound is usually better served by fixing the pages buyers land on after the referral than by competing for AI answers it has no presence in. Referred buyers still check you, and what an engine says about you is part of that check.
Should we wait until AI search sends more traffic before investing?
Waiting for the traffic to arrive means waiting for something that may not arrive in that form. SE Ranking's 2026 study measured AI platforms at 0.32% of all website traffic across 101,574 sites, and found Google sending about 134 times more visitors. The case for the work is influence over shortlists, not referral volume, so judging it on traffic answers the wrong question.
Can we do this ourselves instead of hiring anyone?
Yes, and some companies should. A team with a writer who knows the category, a developer with an hour a week, and somebody willing to run the prompts monthly can do most of the page work. The parts that are hard to do in-house are the measurement discipline and the honesty about what is not working.
What is the smallest sensible way to start?
Run twenty buying prompts across ChatGPT, Gemini, Claude and Perplexity, record who gets named and which pages get cited, and rewrite the five pages that should have been quoted and were not. That costs a week of someone's attention and tells you whether a larger programme has anything to work with.