Organic search clicks have fallen 42% since AI Overviews expanded. That number gets thrown around a lot. What gets mentioned less: in B2B tech categories like cybersecurity, cloud infrastructure, and marketing/sales SaaS, top-of-funnel content actually gained traffic over the same period.

Both things are true at once. The question isn't whether TOFU is dead. It's whether your TOFU is built for the environment it now operates in.

The Split Nobody Expected

A recent analysis of 30 major publishers across nine industries compared August 2025 to August 2026 organic traffic estimates. The results split along a line most content strategists didn't predict: industry type mattered more than funnel stage.

Every domain in finance, healthcare, legal, and consumer tech publishing lost TOFU traffic year over year. Glossary entries, how-to articles, guides. All down. This tracks with Google's YMYL (Your Money or Your Life) categories, where AI Overviews answer the query before anyone clicks.

But cybersecurity and marketing/sales software domains improved. Cloud infrastructure ended up positive collectively. B2B buyers researching a platform migration or evaluating a security vendor go deeper than a single AI Overview can satisfy. Larger budgets, more stakeholders, more consideration time. A paragraph from an LLM doesn't close that loop.

MOFU Didn't Escape Either

The instinctive response to declining TOFU is shifting budget toward middle-of-funnel content. Comparison pages. Product vs. product pages. Feels safer.

The data says otherwise. When the same study compared TOFU and MOFU trends within each industry, the patterns were "remarkably similar." In YMYL verticals, MOFU declined alongside TOFU. In B2B tech, both grew. Funnel stage wasn't the variable. Industry was.

If you're considering a reallocation from awareness to decision-stage content: that move might not sidestep the problem. If your industry is getting compressed by AI Overviews, the compression applies across stages. The fix isn't moving budget down-funnel. It's changing what your content actually does.

What Still Earns Clicks (and Citations)

Three TOFU formats are showing resilience, and they share a trait: AI can reference them but can't reproduce them.

Interactive tools. Calculators, screeners, assessments. One analysis found plain educational how-to content gets cited by LLMs only about 12% of the time. Ungated tools picked up AI Search traffic consistently, with AI summaries recommending specific tools by name. An LLM can write a how-to article. It can't run your ROI calculator for the user.

SME-driven content with a point of view. AI Overview responses are, by design, generic. They won't take a position, critique a vendor, or share a firsthand mistake. Content from a real subject matter expert who's run the process and formed opinions about trade-offs reads differently from an AI summary assembled from five existing explanations. That difference is measurable in engagement and citation rates.

Local and context-specific content. LLMs still lack reliable location awareness. AI Overviews appear far less often for local queries. For companies with geographic service areas, locally tailored content remains a stable acquisition channel.

The Measurement Problem Underneath

Only 36% of marketers can accurately measure content return. That stat should bother you more than the traffic decline. If you can't measure whether TOFU influenced pipeline, you can't make a defensible keep-or-kill decision.

TOFU conversion is structurally weak: roughly 0.19% to 2.3% of visitors become leads, and somewhere between 1 in 1,000 and 1 in 10,000 becomes a customer. Judging TOFU on last-click revenue is like judging a billboard by direct-response rates. The measurement model has to match the asset's role. Instrument pipeline-influenced reporting, assisted conversions, content-to-demo progression. That's directional attribution, not proof, but it's better than pageviews or nothing.

The Allocation Decision

If you're in B2B SaaS, the data supports continuing TOFU investment, with conditions. Build fewer, stronger assets. Lead with original insight or a proprietary tool. Structure content for AI readability (direct Q&A blocks, entity clarity, structured data) so you earn citations in AI answers, not just traditional rankings. And measure downstream influence, not vanity traffic.

If you're in a YMYL vertical, the honest read is harder. Even the largest publishers in finance and healthcare lost ground. If dominating AI Overview citations isn't enough to protect Mayo Clinic or Investopedia, smaller publishers competing for thinner SERP slices face steeper odds.

The scaled AI content play looks worse by the quarter. Claude now watermarks output. Pangram rolled out across Substack. Google's successive spam updates go beyond traditional ranking signals. Fully automated TOFU has little long-term value, even if it still picks up stray clicks today.

The old question was "should we publish more TOFU?" The better question for the back half of 2026 and into 2027: does each page we publish deserve to exist? If the answer requires a pause, that pause is the strategy working.