Ziff Davis VP: Target Bottom-of-Funnel Terms, Not Big Keywords

By Elena Vasquez

Jonathan Berthold, VP of Revenue at Ziff Davis, calls bottom-of-funnel targeting "100% the easiest low lift opportunity" in SEO, because most marketers chase high-volume informational keywords that LLMs, AI Overviews and AI Mode have already absorbed. He laid out his testing playbook in an interview on The Search Engine Journal podcast, covering everything from leadership buy-in to how long an experiment needs before anyone judges it.

Berthold works in both SEO and paid media. His advice targets teams with flat traffic and no budget to fix it.

What should you test first when traffic is flat?

Berthold starts with a full site analysis to find out why traffic stalled. He has inherited sites in that condition before, from B2C SaaS companies to informational lead-generation sites. The analysis decides the fix: improve existing pages, or build new ones to fill a topic gap.

Either way, the experiment targets commercial intent. Top-of-funnel traffic is being "siphoned by LLMs, AI Overviews, AI Mode," he said, and that traffic "is going away and/or doesn't really convert or add to the bottom line." He would test one of two approaches:

• Optimizing existing pages for commercial and transactional keywords.
• Creating new content to fill topic gaps competitors are not covering.

Without budget, Berthold relies on AI for research. He uses it to learn what ranks for bottom-of-funnel terms and what added perspective would make new pages "unique and inviting to search engines like Google."

His core observation: there is "a massive market for bottom-of-funnel traffic that a lot of SEOs sleep on." Keywords above a thousand monthly searches are easier to build strategy around, but Reddit, LinkedIn and Wikipedia already rank for them.

How do you get leadership to approve a test?

Berthold treats approval as a communication problem. He never tells a leader the team will test something and hope for a result. Every experiment starts with a hypothesis, and the pitch explains what the test will reveal.

He frames the ask around long-term payoff. When the pitch says "this experiment will allow us to learn X, Y, and Z so that we can earn this much into the future," he said, it is "easier understood by a leadership team that largely works within P&L."

Why start small before sitewide changes?

A failed sitewide SEO test can destroy an entire channel. When a test goes the wrong direction across a full website, "you tank the whole organic," Berthold said. His rule for SEO and paid media: start with a very small section, then iterate and expand as results come in.

Paid media taught him the same lesson. Tripling a budget in one day "ruins all the optimization," he said, so budgets should rise gradually while the system learns. The SEO equivalent of that mistake is rolling out a schema or content experiment sitewide and tanking pages that already perform.

Results also need replication. A result should repeat "so that it's not just a random result," until a pattern emerges that applies to the rest of the work.

How long should a test run?

Berthold gives an SEO experiment about a month, relying on Search Console data treated as directional. Google needs time to crawl, process and adjust rankings, so "definitely within a week, you won't see results." After 30 days there is usually enough data to see whether impressions move. He waits 60 days at most.

Paid media follows a different rule because feedback is immediate. He either reviews tests every two weeks, or applies a spend threshold of two to three times the target cost per acquisition. Once a test spends that much, the team iterates or kills it.

What test do most marketers skip?

Berthold sees the same gap in SEO and paid media: teams focus on generating traffic and overlook "what happens when a user actually hits your website." Landing page and funnel experiments should always be "an on thing," he said, because conversion rates can keep rising at each funnel stage, "even by a few basis percentage points."

He also uses AI citation data "to unearth link building opportunities." Many companies track which sites cite their competitors, he said, "but not a lot are taking the next step, which is reaching out to these publications." Pages citing competitors on bottom-of-funnel keywords likely rank well and get cited in LLMs, so earning a mention there compounds.

Does search volume still matter?

Berthold no longer puts much stock in it. If context is right, a long-tail keyword can be worth targeting even at one, 10 or zero reported searches a month. His agency example: a composites client selling fiberglass conduits had tiny search volume, but if 10 people searched monthly and two landed on the site, "very likely they're in that market. That's a half-a-million-dollar contract."

As LLMs spread, he expects search volume to be "not fully representative" of real queries. His closing advice: as AI absorbs informational searches, test on the pages where visitors decide whether to convert. A small test there can show a measurable business result within a month or two, and that number makes the next experiment easier to approve.