Does top-of-funnel content still deliver ROI in 2027? It depends on your industry

The question that opens Adam Gnuse's analysis on Search Engine Land is uncomfortable for anyone who produces content: with organic clicks down 42% since the expansion of AI Overviews on Google, does top-of-funnel content (TOFU, the kind of article that answers "what is X" and "how to do Y") still generate enough traffic to justify the effort of writing it?

The short answer that emerges from the data is: it depends almost entirely on the industry you're in. And that changes how Brazilian SEO should prioritize editorial production.

What the data shows

Gnuse pulled organic traffic estimates from Semrush for August 2026 and compared them with the same month the previous year across 30 of the largest digital publishers, spread across nine industries. The methodology is honest about its limitations: third-party panel data doesn't measure precise traffic totals, but it does capture general trends.

The TOFU segmentation was done pragmatically: pages that rank for keywords with "what is..." and "how to...", plus URLs containing "guide". It's not surgical, but it's replicable, and that's what matters for reading the market's movement.

The result splits the map into two opposing territories:

• Lost traffic: all domains in finance, health, legal, and consumer tech. Glossaries, "how-to" articles, and guides fell together. LegalZoom was the extreme case, with a 90% year-over-year drop, attributed by the company itself to changes in Google Search, not to internal changes on the site.
• Gained traffic: all cybersecurity and marketing/sales software domains improved. Cloud infrastructure ended up positive collectively.

The loss pattern coincides with Google's YMYL (Your Money or Your Life) categories, exactly where AI Overviews appear most aggressively. These are the topics where Google assumes the user wants a fast, authoritative answer without needing to click.

Top and middle of funnel move together

One of the study's most useful findings knocks down a common assumption. Gnuse compared TOFU with middle-of-funnel (MOFU) markers: product/service comparison pages ranking for "vs." and "best". The expectation would be that MOFU, seen as more "evergreen", would hold up better.

That's not what happened. TOFU and MOFU moved in the same direction within each industry, rising or falling together. The practical implication is direct: a YMYL brand that cuts its top-of-funnel budget to invest in middle-of-funnel doesn't escape the impact of AI Overviews. The problem isn't the funnel stage, it's the industry.

Why B2B tech holds up

The explanation makes sense from a buying-behavior standpoint. A B2B buyer evaluating a platform migration or a security vendor invests more time in the topic and gives more weight to demonstrated expertise than a YMYL consumer looking up symptoms or a tax rate. B2B decisions involve larger budgets and more consideration, so the research goes deeper than a single AI Overview can satisfy.

There's also an interesting second technical factor: B2B software and infrastructure topics change fast enough that LLM training data can't reliably keep up. When an LLM doesn't have the answer ready, it has more reason to search for and cite the page that does, which can still generate the user's click.

The TOFU strategies expected to work in 2027

Gnuse points to three fronts with positive signals, and the logic behind each one applies to anyone producing content in Portuguese.

Local search that serves local needs. Recent studies show consistency in local search traffic, because LLMs still have unreliable location awareness and AI Overviews appear far less in local queries. The caveat: going beyond programmatic location pages and delivering content that's genuinely useful for each service area.

Interactive digital tools. This is the most actionable insight. In an earlier analysis by Gnuse comparing organic traffic and LLM-referred traffic, educational "how-to" content was cited by LLMs only about 12% of the time, because the models themselves already write that kind of generic content. Ungated tools, though, like specialized calculators and industry screeners, performed the opposite way: nearly every tool page captured some AI search traffic, and summaries would even recommend a tool by name when asked to evaluate something. As the piece sums it up: tools are one of the few TOFU formats that AI summaries can cite but can't reproduce on their own.

Real subject-matter expertise. AI Overview answers are fast, generic, and personality-free by design. A "what is X" article written or informed by a human SME (someone who's run the process, made mistakes, formed opinions on trade-offs) reads differently from a summary assembled from five existing explanations. And tapping that expertise is often the shortest path to original data or insights, which raises the odds of being cited by the AI itself.

The elephant in the room: AI-scaled content

The article doesn't shy away from the most delicate topic. Hyperscaled AI content generation has become standard TOFU practice at some outlets, and it's easy to see why: where AI writing is strongest is exactly where most top-of-funnel content lives, the generic informational article. It's an automated, low-effort lottery ticket.

Gnuse argues that this bet is in trouble due to three converging trends:

1. The quality improvement in AI writing has slowed. It's early to call it a plateau, but few would argue that model-to-model gains in written text haven't slowed down over the past year.
2. Tools are flagging AI text. Claude now watermarks its output, and Pangram has been deployed on Substack. AI detection is gaining traction, and users are starting to expect it as part of the web experience.
3. Google may be changing its mind. Successive updates have been aggressively cracking down on scaled content, apparently going beyond traditional engagement signals, on-page performance, and backlinks.

The conclusion is balanced: there's no reason to avoid AI in writing, since human content enhanced by AI performs strongly, and sometimes the strongest. What seems to have little long-term value is fully automated, highly scaled content.

What changes for Brazilian SEO

The practical takeaway is about prioritization. If your content falls into a YMYL niche (health, personal finance, legal), the recommendation is at minimum to reset expectations: even shifting from top to middle of funnel doesn't solve it, because both fall together. If you work in B2B tech, cloud, DevOps, or cybersecurity, top-of-funnel is still a reliable strategy, especially when combined with interactive tools and real expertise from people who live the problem.

And it's worth repeating the reminder that runs through the whole piece: SERP competition has gotten fiercer. To generate real clicks, content now needs to beat competitors, AI Overviews, and LLMs, all at the same time. The question the author leaves as a final filter is the most useful of all to measure before publishing: does this page deserve to be published?