Why considered software purchases are where this work pays, and which pages carry it.
AEO for B2B SaaS is the work of making a software company quotable by AI assistants during the weeks of research that precede a purchase, so assistants name it in category answers.
A consumer buying a kettle asks an assistant once and buys. A finance director replacing a billing system asks an assistant a dozen times over six weeks, in a dozen different framings, long before anyone books a demo. That difference is the whole argument for AEO in B2B SaaS — the research phase is long, it is now partly conducted inside assistants, and every question asked there is a chance to be named or ignored.
It is the practice of making your software company retrievable, quotable and attributable inside AI assistants, aimed at the questions a software buyer asks during evaluation. Where SEO competes for a position in a list of ten blue links, AEO competes to be the source an assistant lifts a sentence from and names.
The mechanics are the same for any business: the retrieval agent has to be able to fetch the page, the page has to contain a claim that survives being quoted on its own, and the company has to resolve as an entity the model can identify with confidence. The general definition, in plain terms →
What changes in B2B SaaS is the economics. The buying cycle is long enough that an assistant is consulted repeatedly, the deal is large enough that one additional shortlist appearance matters, and the questions are specific enough that a well-written page can actually answer them. Those three conditions rarely appear together, and they are why this category is worth the work when many are not.
Because the cost of being absent compounds over the cycle rather than happening once.
In an impulse purchase, being missing from one answer costs you one sale. In a considered purchase, being missing from the category answer means you are not on the shortlist that gets built in week one — and everything downstream, the comparison questions and the due-diligence questions, is asked only about the vendors already on it. You are not losing a deal at the end. You are being excluded before anyone knew you existed.
There is a second effect specific to software. A SaaS evaluation usually involves several people who did not attend the same meetings: a champion, a technical reviewer, a finance approver, sometimes a security reviewer. Each of them researches independently, each of them now asks an assistant, and each of them is asking a different question about you — integrations, pricing model, deployment, compliance. A company that answers only one of those questions clearly gets named in one conversation out of four.
And when the assistant cannot fetch your site, the description of your product that reaches the buyer is written by whoever the assistant could fetch. In SaaS that is usually a competitor’s comparison page or a review aggregator, neither of which is trying to represent you fairly.
The branded question is “what does Acme do?” or “is Acme any good?”. The category question is “what tools handle revenue recognition for a SaaS business?”. They behave completely differently and confusing them is the most common way a SaaS company misreads its own position.
| Branded question | Category question | |
|---|---|---|
| Asked by | Someone who already knows you | Someone who does not |
| Wins you | A cleaner due-diligence answer | A place on the shortlist |
| Depends on | Entity clarity, consistent descriptions, third-party profiles | Retrievable, quotable pages plus corroboration elsewhere |
| Difficulty | Low — you are usually the strongest source about yourself | High — you are competing with review sites and competitors |
| Reading a win as success | Misleading. Almost every company wins these. | Meaningful. This is the growth number. |
Most SaaS companies test their visibility by typing their own name into ChatGPT, getting a decent answer and concluding they are fine. They have tested the easy question. The one that produces pipeline is the category question, and it is the one where you are competing against G2 listings, “best tools for” roundups and your competitors’ comparison pages. Track them separately or you will flatter yourself. How to track AI citations without fooling yourself →
Not the blog, in the first instance. The pages that get cited in software answers are the ones that answer a specific, checkable question about your product, and on most SaaS sites there are four of them.
| Page | Answers the question | What makes it quotable |
|---|---|---|
| Pricing | “How much is it, and what do I get?” | Named tiers, stated prices, what is included and excluded, in text |
| Comparison | “How does it differ from X?” | A table of concrete differences, with the cases where the other tool wins |
| Integrations | “Does it work with our stack?” | A named list, one page per major integration, describing what the integration does |
| Use case | “Does it work for a team like ours?” | The segment named explicitly: role, industry, company size, the job done |
The common thread is specificity. “Flexible pricing to suit teams of every size” cannot be quoted by anything, because there is no claim in it. “The Team plan is billed per seat per month and includes SSO and the audit log” can be lifted straight into an answer, and frequently is.
“Contact us for pricing” removes you from every pricing answer. Buyers ask assistants what things cost, constantly, and when your page has no number the assistant answers from whatever it can reach — a review site’s estimate, a forum post, a competitor’s comparison table. You do not avoid the pricing conversation by hiding the price. You hand authorship of it to someone else. If you cannot publish a full price list, publish a starting price, a model and what changes the number.
Yes, and they are the highest-leverage page type in this category, because comparison questions are asked constantly during evaluation and there is very often no honest page answering them.
The rule that makes them work is counterintuitive: the page has to be genuinely fair. A comparison page that concludes you win on every axis is marketing, reads as marketing, and is a weak source for an assistant to lean on. A page that states clearly where the other product is the better choice is a far stronger source, and it has a second effect — a buyer who reads an honest disqualification and stays is a much better lead.
What a comparison page needs: a table of concrete, checkable differences rather than adjectives; a plain statement of who each product suits; a dated note on when the comparison was last verified, because competitor features change; and no claim about the other product you cannot support. Write it as though their team will read it, because they will.
Publish one page per significant competitor rather than one page covering all of them. A single page mentioning six competitors cannot be cleanly retrieved for any one of those comparisons.
Start narrow. A small number of precise pages outperforms a large number of vague ones, and a pre-launch or early-stage SaaS company is often better placed than an incumbent because there is nothing to unpick.
llms.txt and your footer. Models quote these strings close to verbatim.Six pages done to this standard will be retrieved and quoted more often than sixty pages of blog posts about industry trends. Volume is not the lever here; extractability is. The full checklist, ordered →
Measure on category and comparison prompts, run repeatedly, under fixed conditions — not by asking once and screenshotting a good answer. Assistant output is non-deterministic, so a single favourable run proves a citation is possible, not that it is reliable.
The practical setup for a SaaS company: ten to twenty prompts drawn from real discovery-call language, weighted towards category and comparison; three runs per prompt; logged out, with the market pinned to where you sell; citation share and mention share recorded in separate columns; and a column for who was cited instead of you. Re-run the identical set monthly.
Referral traffic is a secondary signal at best, because a meaningful share of assistant traffic arrives with no usable referrer. Treat it as corroboration, not measurement. The number that matters is whether you are named in the category answer more often than you were last month.
There are real cases, and pretending otherwise would be dishonest.
The second and fifth are the ones worth being blunt about. The longer case for and against → and what engagements cost →, so you can judge it against the alternatives rather than against zero.
robots.txt; Organization and Product schema corrected; one company description agreed and applied everywhere; baseline measurement taken on the frozen prompt set before anything else ships.Nothing in that sequence guarantees a citation, and anyone who tells you otherwise is selling something a model cannot deliver. What it does is remove the reasons you are currently uncitable, in the order that matters. If you want to know which of those reasons apply to your site before committing to any of it, the free AI visibility check tests access per agent, reads your key pages the way a retrieval agent does, and reports who is named in your category answers instead of you.
They overlap and they are not the same. SEO competes for a ranked position; AEO competes to be the source an assistant quotes and attributes. Crawlability and clear structure help both, while entity consistency, self-contained sentences and retrieval-agent access matter specifically for AEO.
Pricing, one comparison page and one use-case page, after crawler access is confirmed. Those three answer the questions buyers put to assistants during evaluation, and they are usually the pages written most vaguely because they were written to sell rather than to be quoted.
They are among the most retrievable pages a SaaS site can publish, because comparison questions are asked constantly and honest answers are scarce. A page that concedes where the competitor is stronger is a better source than one that claims a clean sweep, and it converts better too.
Yes. Buyers ask assistants what software costs, and with no number on your page the assistant answers from review sites, forums or a competitor's comparison table. Publishing a starting price and the pricing model keeps authorship of that answer with you.
Longer than a campaign and shorter than a rebrand. Access fixes take effect as soon as agents re-fetch; page rewrites need re-crawling; corroboration from third-party sources builds slowly. Measure monthly on a frozen prompt set rather than watching for a sudden change.
Often yes, provided the product's audience is settled, because there is no legacy content to unpick and six precise pages can be shipped quickly. It is not worth it if positioning is still moving, since every page would be rewritten before it was ever retrieved.