How to audit your B2B SaaS positioning for the AI era

LinkedIn stopped optimizing for Google in 2025.

Their non-brand content traffic fell away while rankings held steady. So they retired the traditional SEO playbook, stood up an internal AI Search Taskforce, and rewrote their guidance around getting quoted inside AI answers. Their KPIs now count mentions and citation frequency rather than sessions.

When the platform that taught B2B content marketing quietly abandons its own playbook, that's worth a morning of anyone's attention.

Underneath it, the thing that actually changed is simpler. Buyers stopped researching the way they did eighteen months ago. They ask an assistant, get a synthesized summary of a company's entire digital presence, and form a view long before they load the homepage.

---

The secondhand pitch

Most B2B positioning still assumes a buyer who finds us through search, arrives at sales still open-minded, and forms an impression from our website and our deck. Every part of that has moved.

6sense's 2025 survey of nearly 4,000 B2B buyers found 94% had already ranked their shortlist in order of preference before contacting any seller.

The demo, the RFP, the POC. Those are validation exercises now.

So our positioning reaches buyers as a paraphrase. A machine assembles it from our site, our review profiles, a competitor's comparison page, and a Reddit thread from last spring, then hands the buyer a tidy summary.

Call that the secondhand pitch. It's the version of us that gets heard.

Nobody on our team wrote it, approved it, or has read it. Whatever we didn't put down plainly, the machine fills in from somewhere else, and it fills it in fluently.

---

Reading it back

The first audit takes thirty minutes.

List the fifteen to twenty questions a buyer asks while evaluating our category. Stay above the vendor name. We want the problem space, the class of tool, the questions that come before anyone types a brand.

For an API monitoring company that's "what's the best API monitoring tool for mid-market," or "how do engineering teams cut incident response time," or "what should I look for in an observability platform."

Run each one through ChatGPT, Perplexity, and Gemini. Write down three things.

Do we appear? If we do, where in the answer and in what company. If we don't, who's there instead and how are they described.

Is the description right? Plenty of teams find the machine describing features they retired, or capabilities they never shipped. LLMs prioritize fluency over accuracy, and a plausible wrong answer competes directly with our messaging.

What words does it use for our category? That framing accumulated out of what the market has been saying about our space, which tells us what we're really being compared against.

Run it quarterly, because it moves.

The companies that show up consistently have two things going for them. Their positioning is specific enough that a machine can tell which question they answer. And their brand is corroborated somewhere other than their own site, across review platforms, analyst coverage, and industry press.

Gartner calls that second one entity authority. A domain authority score has nothing to do with it.

---

Claims three competitors could make

Now audit the positioning itself.

Most B2B SaaS messaging is a capability argument. Here's what we do, here are the integrations, and surely that justifies the spend.

That argument is losing. AuditBoard's win-loss interviews found features had almost nothing to do with why buyers chose them or passed. What decided it was ROI confidence, and buyers had to believe the investment would pay back and that they could show a CFO the math.

The bar moved from "does it do the thing" to "will it demonstrably improve the business, and can we measure it."

So read the homepage and ask what business outcome a stranger could name after thirty seconds on it. If they'd need a case study, a demo, or a sales call to answer, the positioning is doing too little of the work.

Then take the harder pass. Every primary claim on the homepage, the value proposition, the product description, the category pages.

For each one, ask whether three competitors could put the same sentence on their site tomorrow. A yes means the claim establishes category membership and nothing beyond it.

Investors have started publishing what they've stopped funding in AI pitches, and the lists from Emergence Capital, AltaIR, and F Prime name the same tells. "AI-powered," UI as the differentiator, integration depth as a moat, "purpose-built for X" with no sharp ICP behind it, thin workflow layers over somebody else's model.

Buyers stopped believing those phrases around the same time the money did.

And there's a question every buyer now asks, usually not out loud. Why can't we just build this ourselves with AI?

Retool's 2026 Build vs. Buy report found 35% of teams have already replaced at least one SaaS tool with a custom build.

Our positioning needs an answer before sales walks into that question, and "we're better" won't hold.

What holds is everything a weekend build doesn't come with. Compliance posture, the security review, the maintenance nobody budgets for, the years of institutional knowledge sitting inside the tool, a roadmap somebody else is funding.

That's a story about what the buyer stands to lose. A feature list can't tell it.

---

What a machine can quote

Right message, wrong format, no citation.

LLMs pull from content that answers a question outright. Set a three-thousand-word essay on the future of observability next to a page that names the best API monitoring tool for mid-market teams, says which capability makes it that, and cites a source.

Nobody would call the second one good writing. It's the one that gets quoted.

Which argues for two layers in the content library. The opinionated long-form that builds a brand with human readers, and a structured layer that feeds summaries.

The structured layer means FAQ pages that answer category questions directly, comparison pages against named alternatives including build-versus-buy, "who this is for" pages by persona, and schema markup so there's a verified source for the basic facts about the company.

Then there's the part no dashboard sees. Buyers form opinions in Reddit threads and private Slack groups, and none of it reaches attribution.

Search the company name plus Reddit and read what comes back. That's raw material for the secondhand pitch, and it's what buyers read before they visit the site.

If people who don't work here describe our category and our product clearly, the machine has something accurate to work from. If that presence is thin, it guesses, and it guesses fluently.

---

The teams doing this now are building an awareness moat that budget and reach can't override. They're the ones an assistant names when a buyer asks for a shortlist, described accurately, as the obvious answer for a defined audience.

Somebody is already pitching for us.

Go read what they're saying.

---

What to do next

If you've never seen how AI describes your product, start there. Fifteen questions, three assistants, one morning. Whoever appears instead of you is your roadmap.

If what comes back is accurate and still sounds like everybody else, that's a positioning problem, and no amount of content structure will fix it. A positioning audit is built for exactly that.

If that's where you are, start here. The first conversation is free.

---