AEO for B2B SaaS: be in the answer before the demo.

AEO for B2B SaaS is the work of getting your product named when a buyer asks an AI assistant for a shortlist — “best CRM for a 12-person sales team” — before they visit a website or talk to sales. More than half of B2B software buyers now start vendor research in an AI chatbot more often than Google (G2 buyer research, 2026). The reply names two to four products, the demo list gets drawn from it, and no tool in your funnel records the event. What follows: the questions stage by stage, who wins them today, and what contesting them costs.

The pipeline stage no CRM logs

SaaS go-to-market runs on instrumented stages — first touch, MQL, demo booked, closed-won — each with a dashboard and an owner. The stage that now sits in front of all of them has neither. A buyer describes team size, stack, and budget to an assistant and asks what to look at. The reply is a shortlist with reasons attached, composed in seconds, arriving before your site logged a visit.

Surveys of what happens inside that chat point the same way: 72% of B2B software buyers report using ChatGPT to evaluate vendors, and 69% say AI guidance changed which vendor they chose (buyer survey reported by MarketScale, 2026). Vendor-adjacent surveys, not a census — but nobody has published data pointing the other way.

What separates this stage from every other leak in a funnel is the silence. A lost deal gets a loss reason; a dropped trial leaves a churn row. A shortlist you never entered leaves nothing: no impression report, no console, no alert from any engine. The question was asked, the answer named other vendors, and your quarter just reads as slow. Nobody reports the absence to you — it has to be discovered.

The shortlist math: fewer seats than you think

The same data set describes ChatGPT and Perplexity typically surfacing two to four brands per category question (AuthorityTech, 2026). Read the numbers together: the list buyers carry into demos is converging on the size of the answer engines compose. If the category answer has three seats and incumbents hold them, the arithmetic for everyone else is a demo that never gets booked — multiplied by every buyer who asked.

The counterweight is churn: 40–60% of the sources AI engines cite change every month (eMarketer, 2026). Seats reopen constantly — for whoever has the quotable page standing when they do. That cuts for challengers, which is rare among marketing channels.

What SaaS buyers actually ask, stage by stage

Fourteen question shapes from published 2026 buyer research, grouped by funnel stage. The vendor names are invented — Ardent CRM and its rivals are this library's fictional example category — but the shapes are the thing to study, because each stage is decided by a different kind of page.

One more property changes how you measure: the engines do not agree with each other.

Who owns your category's answers today

Run your own head term and read the citations before you believe any vendor's chart, including ours. The pattern repeats across SaaS categories: for “best X” questions, engines reach for pages built to answer list questions — aggregators' category rankings, comparison listicles, community threads — because those pages name several vendors and state tradeoffs in sentences that survive being quoted. Vendor sites get cited mostly for their own brand's questions: pricing, docs, “is it legit.” The category answer — the one that forms shortlists — routes through pages you do not control.

Two consequences. First, your review-site profiles are doing answer-engine work whether or not anyone maintains them; stale ones work against you on a surface you are not watching. Second, community threads carry outsized weight in exactly the head-to-head and migration stages where deals are won — why AI cites Reddit so much covers the mechanism. And when a rival holds the seat you wanted, the forensic method in competitors showing up in AI answers shows how to read which citations put them there, and which positions are genuinely contestable.

The page an engine can quote is the honest one

Aggregators do not win category answers because engines favor aggregators. They win because a “best X for Y” question can only be answered by a page that names several products and commits to tradeoffs — and most vendor pages refuse to do either. That refusal is the opening: publish the honest version yourself, naming the rivals, conceding the rows they genuinely win, stating your own facts precisely enough to check.

The honesty is what makes the page usable: an engine assembling an answer needs sentences that survive being lifted out of context. “For enterprise reporting, the incumbent is the sharper tool; for a 12-person team the math favors X” survives. Unqualified self-praise does not. We run this play ourselves — our own comparison pages, like AEO Action vs Profound, hand entire rows to the product being compared.

The spec stage rewards a second shape: docs-shaped answers — the plan fact stated like documentation rather than persuasion. “The API is included from the $49 tier; the SOC 2 report is available on request.” Pages that hide those facts behind a demo form cannot be quoted for the questions that mention them. Structured data extends the same idea to machines: Product and FAQ markup state prices, tiers, and integrations in a format a machine can parse without interpreting, and they still drive classic rich results. Google states no special markup is required for AI Overviews or AI Mode — treat schema as precision, not a lever. The playbook for the biggest surface is in ChatGPT SEO.

What the department builds for a SaaS company

AEO Action's bet: the scarce thing in SaaS marketing is not knowing any of the above — it is the distance between knowing and shipping. So the product runs as a department. The scan re-asks your buyer questions daily — 50 on Solo, 100 on Solo Pro, regenerated from your business profile each run — across up to six engines: ChatGPT and Gemini through the live consumer interface, Perplexity through its API, Google AI Overviews from live results, Copilot from Bing's answer layer, Claude in a labeled model mode, never counted as live. Competitor pressure is stored as its own score, so “who took the seat” is a number that moves, not a vibe.

Each day the orchestrator picks the highest-value move from the scan and prepares the full artifact into an approve queue. For a SaaS company:

• Comparison pages— the honest head-to-head for the “X vs Y” asks a rival or an aggregator currently answers for you.
• Use-case landing pages— “best X for Y team”-shaped, aimed at questions the scan found you absent from.
• FAQ pages and schema — the docs-shaped facts in visible text and in matching Product and FAQ markup; what the markup states always matches the visible page.
• A Reddit packwhen community threads hold your category's answers — prepared for you to participate with, not to astroturf.

Nothing publishes without your Approve. When you click it: Webflow and Squarespace publish directly; on WordPress the plugin publishes the approved page as a real page on its hourly sync; a custom-built marketing site — common in SaaS — gets paste-ready HTML and schema, verified once your team ships it. The URL then goes to IndexNow within minutes; submission is our half, indexing is Bing's decision. Why bother: independent testing found roughly 87% of ChatGPT's cited pages also rank in Bing's top results (Seer Interactive, 2026) — a correlation measured from the outside, not a pipeline OpenAI documents, and overlap enough to be worth working. The full epistemics are in Bing and Copilot optimization.

Then the part most tools skip: proof. Every approved landing page is registered against its target question and re-checked on each scan for up to 120 days — a dated receipt, shipped date A to first-cited date B. A plain-English work report lands every 7 days. And the AEO Pixel plus a GA4 integration read the assistant-referred visitors actually reaching your site — where GA4 alone goes blind is covered in tracking AI traffic in GA4.

Three ways to staff it, one ledger

A SaaS team has three real options here, and the honest comparison is what each monthly number actually contains. One disclosure before the table: we build the product in the highlighted column. Judge accordingly.

The retainer is the right buy for some teams — the cases are argued honestly in what AEO services actually charge. Still deciding whether the discipline deserves budget at all? Start with the skeptic's version: is AEO worth it.

When AEO Action is the wrong buy for your SaaS

Five real disqualifiers, stated the way we would want a vendor to state them to us:

• Your marketing site is a custom build and you expect a push. Direct publishing reaches Webflow and Squarespace; WordPress gets the plugin. A Next.js or fully custom site — the most common SaaS setup — gets finished, paste-ready HTML and schema, verified once live. Real output, but an engineer on your side merges it.
• You already staff product marketing at scale. If a PMM and writers cover comparison and use-case content well, our execution layer duplicates people you employ. Buy a cheaper daily tracker and spend the difference on their hours.
• You need enterprise analytics depth. Our ceiling is 100 prompts per daily scan on Solo Pro. Programs running thousands of prompts with panel-scale query volumes belong on enterprise platforms.
• Procurement needs SOC 2 or an SSO checklist today. We do not have them.
• You want to enter a card and be billed automatically today. Card checkout is not live yet, and neither is scheduled email delivery — reports live in the product and export as PDFs. The trial genuinely takes no card, which is the same fact read from the other side.

The cheapest fact in your funnel

Everything above compresses into one testable question: when a buyer asks for your category's shortlist, are you in it? The free audit answers that in about three minutes — 10 buyer questions across ChatGPT, Gemini, and Perplexity, 30 live answers, the three most revealing quoted word for word, plus a 30-day plan. If the answers already name you everywhere, close the tab; you did not need us. If they name your competitors, you know which seat to contest first.