The Content Funnel Is Dead. Stop Investing in TOFU Like It’s 2019.

Finding 1: Winning Early Doesn't Guarantee Winning Later

The old model said a brand that shows up early should be more likely to show up later. We found the reverse, and it was consistent and strong.

Compared with brands that did not appear on the educational question, brands that did appear were named less often later, not more.

The reason is structural, not mysterious. Each stage of the journey surfaces a different kind of page. Educational questions pull up vendor-neutral explainers and definitions. Comparison questions pull up “best tool” lists, review sites, and product pages. Purchase questions pull up pricing pages, case studies, and analyst rankings.

Different pages, different brands. The overlap is small by design. Put a number on that overlap. Of the brands named at the comparison stage, 88% were never named at the educational stage for the same topic on the same engine. Only 146 of our 4,654 records were named at all three stages.

That also explains something in the next section that looks contradictory at first. Naming rates rise sharply as buyers move down the funnel, from 14% to 47%. Far more brands get named later. They are just mostly different brands, so the pool grows while the overlap stays small.

And substitution is a tendency, not a wall. Of the brands that were named early, 53% were named again later. They are simply outnumbered by the brands arriving fresh at the comparison stage.

In plain terms: owning the “what is customer success” answer does not pre-position you for “best customer success software for mid-market SaaS.” That is a separate contest, and most brands have not entered it.

Finding 2: The Naming Happens at the Comparison Stage

Brands get named much more often once buyers start comparing options. Across the whole dataset, brands were named about 14% of the time on educational questions, then roughly 47% at comparison and 43% at purchase.

That jump is largest on the tools that barely name anyone early on:

On Perplexity, Claude, and Gemini, a brand can be nearly invisible on "what is X" questions and then get named far more often on "best X" questions. Gemini names brands 9 times as often at the comparison stage as at the educational stage. Perplexity 4.2 times, Claude 3.7 times. Google AI Overviews, which names brands most readily early on, shows the smallest jump at 1.6 times.

If your AI visibility dashboard only tracks educational questions, you are not seeing most of your visibility. On Gemini, you would be seeing roughly a ninth of it.

Stage-split tracking is the fix, and it has to run per engine. That means mapping your tracked prompts to educational, comparison, and purchase, then reading each separately. Collapsing them into one number hides the pattern this study found. It is the reason our own platform splits prompt tracking by stage across all five engines.

Finding 3: Challenger Brands Fall Behind After TOFU

Early on, brand size barely matters. Later, it decides almost everything. We sorted brands into four tiers by Domain Authority, from the biggest established leaders down to the challengers.

On educational questions, every tier lands in a tight band of roughly 11% to 16%. Challengers actually edge out the leaders, 15% against 14%. The AI gives them a fair shot. But at the comparison and purchase stages the established tiers climb to roughly 45% to 52%, while challengers stall at 25% and then 18%. The gap is not about content quality early on. It opens up downstream, where being a known brand compounds. Those two challenger figures rest on 163 records, 3.5% of the dataset, so read them as a clear direction rather than precise rates.

One note on method. The tier split is descriptive. When we put domain authority into the regression as a continuous control it was not significant on its own, with an odds ratio of 0.79 and p of 0.43. The tier comparison is the more reliable read of the two.

For a challenger, the practical takeaway is that a healthy early-stage naming rate is not a sign you are winning. It is the one stage where the playing field is level. The work is downstream.

The Playbook: Invest by Stage, Not by Habit

Five moves take you from diagnosing where you currently get named to tracking each stage as its own number.

Step 1: Pick the Stage You Want to Be Named At

Owning the educational answer for a topic does not carry you to the comparison or purchase answer. To find out where you currently stand, take one of your topics, run all three questions on all five engines, and count where you appear. If you show up early and not late, you are in the pattern this study describes.

Step 2: Split Your Strategy by Engine, Not Just by Stage

Google AI Overviews names brands most readily on educational questions, at 26%, so citation work pays there. ChatGPT is different. On four of the five question types we tested it answered educational questions from memory rather than searching, so new pages on your own site will not move it. Only presence in the sources it trained on will. Perplexity, Claude and Gemini barely name anyone early, then name brands 4 to 9 times as often at comparison, so build for that stage.

Step 3: Focus on Building Comparison-Stage Assets

That means brand vs. competitor pages, best tool guides written for one specific use case, clear feature breakdowns, and honest pricing pages. These are the pages AI pulls from on comparison and purchase questions. Better educational explainers will not fill this gap.

Step 4: If You Are a Challenger, Treat the Comparison Stage as the Real Battleground

You will not out-authority the incumbents overnight. What you can do is be present where buyers compare options, which means review sites, community threads, and category guides that name you in a specific use case. Most of that work depends on someone else agreeing to include you, so treat it as sustained effort rather than a campaign with a delivery date.

Step 5: Measure Each Stage Separately

A 14% naming rate on educational questions and a 47% rate on comparison questions are both normal, but they are different conversations. Collapse them into one AI visibility number and you hide the pattern in this study. Split your tracked prompts into educational, comparison and purchase, then read each one per engine.

Steps 1, 2 and 5 are measurement problems. Steps 3 and 4 are production problems, and they are where most teams stall, because comparison assets are a volume commitment and third-party presence depends on other people. That is the work our Visibility and Leads guaranteed outcomes programs are built around.

What the Data Is Actually Saying

The content funnel was a useful model for an era when every piece of content competed in one ranking system and authority built up over time. AI Search does not work that way.

Each stage of the buying journey draws from its own pool of pages and brands. Educational content competes in one pool, comparison content in another, purchase content in a third. Winning one does not automatically win the others.

This is not a reason to stop investing in content, and it is not a reason to stop investing in educational content. Educational pages earn educational mentions perfectly well. What they do not do is pre-position you for the comparison and purchase answers on the same topic.

It is a reason to invest at the stage where you actually want to be named, not the stage that is easiest to write. Most B2B SaaS companies default to educational content because it is easy to produce and shows up in traffic reports. Comparison content is harder and less shareable. It is also where brands get named at 47% and 43%, in the moments that decide who gets bought.

The funnel did not disappear. It broke into three separate markets, each with its own entrants and its own rules. The companies that learn to compete in all three, instead of assuming a win in one carries over, are the ones building for how AI Search actually works.