The title of this post makes a bold claim, so let me be precise about what I actually mean before I cash a check I can’t back up.
Think about a typical content strategy process. You’re doing persona work, buyer journey mapping, understanding what your ideal customers are struggling with: none of that expired. If anything, it matters more now than it did two years ago.
What expired is the assumption sitting underneath all of it: that publishing on more topics is the safer bet.
For about fifteen years, B2B content strategy was built to win one contest: a human typing a query into a search box, scanning ten blue links, and clicking one. Persona development, keyword research, content pillars, the editorial calendar all optimized for that exact moment. Hedge your bets, cover the category, show up for as many searches as you can.
More topics meant more chances to get found.
That process was never built for the second contest now running in parallel: an AI system synthesizing across everything it’s seen about you to decide whether to recommend you at all.
The second contest still rewards comprehensiveness. Real, deep, corroborated coverage of a topic genuinely wins there. If you could cover every problem your buyers have with that level of contextual depth, you should.
Most B2B companies can’t.
A marketing team of three with a budget that doesn’t stretch to fifty topics isn’t choosing between comprehensive and not comprehensive. It’s choosing between comprehensive on a few problems or thin on all of them. That’s the constraint nobody’s calling out.
Fifty scattered topics covered at a shallow depth gives you fifty weak signals. Five topics covered consistently, with a unique solution and deep corroborating evidence, gives you five strong ones.
Contextual depth, uniqueness, and corroboration are exactly what an AI system is weighing, and those five strong signals are what get surfaced.
So the real strategic move isn’t creating more content for the AI era. It’s deciding, on purpose, which problems you’re willing to NOT cover, so the ones you keep can actually win.
This is the crux of Problem Ownership: narrowing your strategy and execution so that you build an evidence layer on a small set of problems that sends a loud, unmistakable signal to the LLMs.
The stakes for getting that call wrong have changed shape. Under the old contest, failure looked like mediocre Google rankings. Page two instead of page one. Annoying, recoverable, an optimization problem you’ll fix next quarter.
Under the new one, when there’s no consistent, corroborated story for AI to retrieve and cite on a given problem, you don’t lose the contest. You’re just not in it.
Let that sit for a minute before we get into mechanics.
Where the Old Content Strategy Process Falls Short
Let’s start with the typical process most content practitioners have deployed for the last 15-ish years (not always in this order):
- Foundational strategy documentation (ICP, personas, UVP, positioning)
- Content audit
- Buyer journey mapping
- Keyword research
- Content pillars
- Channel mapping
- Editorial calendar
- Production workflows
- Measurement against rankings and traffic
I’ve run this basic process more times than I can count, and it’s produced real results for real companies.
What this process doesn’t include, because the question didn’t exist until recently, is a check against AI synthesis. Have we built a comprehensive narrative around a specific set of problems we solve, or are we covering too many topics and diffusing our ability to be cited for any of them?
Without that narrowing step, even a disciplined B2B content strategy can spread its evidence too thin to get cited for anything specific.
What Actually Changes
Here’s roughly how the new sequence runs.
It starts in the same place the old one did. ICP, personas, positioning, the foundational work. That part hasn’t changed, but “fuzzy” isn’t going to cut it any longer.
Foundation Strength
What’s changed is how much slack the system gives you on that foundation. A loosely defined ICP with basic personas and a list of keywords used to still produce something that looked like a working content engine, just a mediocre one. The old contest was forgiving of imprecision because it rewarded volume.
A real ICP defines the specific attributes that make a customer ideal: industry, size, role, constraints, the things that make them them. That context isn’t just useful for conversion. It’s the same context a real buyer brings into an AI conversation when they describe their problem.
Skip that precision and your content speaks in generic terms about a generic buyer, while the actual buyer is describing their situation in specific terms an AI system is trying to match against something. Generic doesn’t match specific, and it doesn’t get cited.
So before anything else changes, that foundation has to actually hold weight. Which is where the real new steps start.
Revenue & Product Anchoring
To begin narrowing your focus into buyer problems, you have to know where you actually win. Name your strongest product/service offerings – the ones with real margin, a sales cycle that doesn’t drag, and customers who stick around and expand.
The offerings that make you the most money have an ideal buyer. Each offering solves specific problems for that buyer. Those problems get your list started.
If you skip this product anchoring step and go straight to buyer problems, you risk picking problems that sound strategically attractive but aren’t grounded in anything you can commercially deliver. You end up trying to own a problem you don’t have the right to own yet.
Ownable IP Inventory
Your existing product or service offerings are likely IP you already own. Do they come with something proprietary, like a named methodology, a branded framework you built, or coined language you’ve published? Or are you just describing what everyone in your category already says, in slightly different words?
That answer changes everything downstream. When you have named, published IP, you’ve got the raw material for differentiated answers that stand out from competitors.
If you’ve got something real but never named it, naming and publishing it is the highest-leverage move available to you, faster and more durable than just producing more content.
And if there’s genuinely nothing proprietary there, you’re competing on signal volume alone, which is slower, more expensive, and rarely wins.
Problem Inventory
Now you actually build the list, and it comes from three things you already have in hand. Your strongest offerings. The ownable IP attached to each one, or the lack of it. And the specific problems your ideal buyer has that those offerings actually solve.
Put those together and you get a real problem list, not a guess. Then translate it into the language your buyers actually use, not your sales deck’s. “I’m really struggling to figure out which marketing channels are generating the best leads for us” lands very differently than “lack of marketing attribution,” and only one of those is what a real buyer types into ChatGPT at 11pm.
Problem Filtering
You likely don’t have the budget or bandwidth to build a comprehensive footprint for every problem on that list. Decisions have to be made. That’s exactly what this step, the part most teams skip entirely, is for: running that list of problems through what we call the Winnable Problem Filter.
Three questions, in order:
- Is this a core problem for your ideal customer and does it align with what you want to be known for?
- Do you have a realistic shot at building enough corroborated evidence to be cited for it, given who already owns the conversation?
- If you were the recommended solution for that problem, would that generate qualified opportunities?
I won’t walk through the scoring here, that’s a longer conversation and would require a demo of our AI360 tool, but there’s a big difference between having a long list of questions/queries we could cover, versus a narrow, strategically aligned list of problems that we can actually win.
Now, the next big change is a step that, in my experience, almost nobody is doing yet, and it’s the one where a specific toolset comes into play.
Evidence Audit: The Audit You’ve Never Done
Content audits have been a staple of content strategy for decades. They check what exists, what’s stale, what’s missing, what’s ranking.
In the AI era, you need something different: an evidence audit.
For each problem you’ve decided to own, is there enough digital evidence out there to support you as the best solution? Not just what you’ve published. Third-party mentions, reviews, partner content, podcast transcripts, anything else floating around with your name attached to it.
AI doesn’t lean on one authoritative source the way search engines weighted backlinks. It builds a picture by synthesizing across everything it’s encountered, which is why an evidence audit has to look at the whole footprint, not just your own content. I covered this in more depth in my last post on AI evidence strategy.
A 2026 study (preprint) by Arxiv coined a useful term for what happens when that footprint is too thin: the “Existence Gap.” Researchers found that a brand can have a better product, better pricing, and happier customers, and still be functionally invisible in AI-generated answers if it doesn’t have enough AI-visible content tied to it.
The study’s strongest evidence came from comparing how differently AI models trained on different language and cultural data recommend the same company, but the underlying mechanism is the same one driving this audit: thin footprint doesn’t get you a worse outcome. It gets you no outcome.
Most companies have never asked what the internet, in aggregate, actually says about them on the problems they want to own. Zero leadership teams can answer that off the top of their head. That gap is the audit’s entire reason for existing.
Evidence Architecture
Once you know what your evidence actually says and the gaps you need to fill, you move into evidence architecture, deliberately designing your plan to build the corroboration layer you’re missing, on the specific problems you’ve decided to own.
AEO Execution
Then we get into execution: the editorial calendar, the capacity, the owned/earned/paid mix, who’s pitching what to whom. I’m not going to build that out here. It’s genuinely specific to each company’s situation.
But there’s one piece of the AEO execution mechanics worth naming precisely, because it’s the most concrete, least obvious thing in this whole post, and getting it wrong is easy.
Links Still Matter. Just Not Like They Used To.
For decades, search engines decided who to trust by counting and weighting third-party links, each one a vote that added up across the web. The anchor text mattered too: a link reading “best CRM for small business” sent a stronger signal than “click here,” and SEO teams spent years engineering exactly that.
AI systems don’t work that way. They’re not counting links or analyzing anchor text. They’re retrieving and synthesizing chunks of text from across a lot of sources to generate an answer.
That does not mean links don’t matter. A link from a credible source is still corroborating evidence, still a mention that counts toward the picture. But the sharper signal is proximity: your brand name showing up repeatedly, near the specific problem language you’re trying to own, across multiple independent sources.
A backlink from an authoritative source rarely contains your buyer’s actual problem language. A good customer review almost always does – your brand name, the problem in plain words, the context they were in. That review is worth more to an AI system than the backlink ever was.
Measurement Changes
Last piece, and I’ll be honest about where this one stands. Measurement changes too, from rankings and traffic to something closer to Problem Ownership tracking.
Are you actually getting cited when buyers describe the problem in their own words? I’m not going to pretend the industry has this fully figured out. It doesn’t. We’re all still building the instruments for a contest that’s only a couple years old.
Permission to Focus
So here’s where I’d leave you.
The paradox we started with: comprehensiveness still wins. Most B2B companies just can’t afford to build the required digital evidence layer on every problem, question, or topic that is relevant to your buyers. There is an actual strategic conversation to be had to decide which problems you’re going to focus on, and give yourself permission to walk away from the rest.
Most teams have personas. Most have a content calendar. Almost nobody has actually run an evidence audit on the handful of problems they’re trying to own.
Sit with your own process for a second and find the step that’s missing. That’s usually the one worth doing first.
P.S. We’re here if you need help executing any of them, or all of them. Give us a shout.
Common Questions
Should B2B companies try to cover every topic their buyers care about, or focus on fewer?
Comprehensive coverage still wins with AI systems, but most B2B teams don’t have the budget to build deep, corroborated content on every topic. The real move is choosing a small number of problems to own completely rather than spreading thin coverage across many. Five topics with real depth and evidence outperform fifty topics covered shallowly.
Do backlinks still matter for AI search visibility?
Yes, but differently than they did for traditional SEO. AI systems don’t count links or weigh anchor text the way search engines did. What matters now is proximity: your brand name appearing repeatedly near the specific problem language you want to be known for, across multiple independent sources.
