流量持平、零預算：低風險 SEO 測試框架

Flat organic traffic does not automatically call for a larger content budget or a major site overhaul. In an interview published by Search Engine Journal, Jonathan Berthold, VP of Revenue at Ziff Davis, outlined a more controlled approach: diagnose the site first, focus on the pages closest to revenue, and treat changes as small learning experiments rather than guaranteed growth projects.

The approach is particularly relevant as AI-generated search experiences absorb some informational, top-of-funnel demand. Traffic volume still matters, but a flat traffic graph can obscure where a site is gaining or losing commercially meaningful visibility.

Start with diagnosis, not a list of tactics

Berthold's first recommendation is a comprehensive site analysis. The aim is to identify where performance has stalled, where topic coverage is incomplete, and which pages or funnel stages have the clearest relationship to conversion.

That matters because “more traffic” is not a sufficiently precise testing goal. A decline in visits to broad informational pages may require a different response from weak visibility for product, category, comparison, pricing, or lead-generation pages. Before changing content, teams should establish what is actually flat:

• Impressions, rankings, clicks, or conversions.
• A specific topic cluster, page type, or market segment.
• Existing commercial pages, missing commercial content, or the journey between them.
• Search demand that still reaches the site versus demand increasingly answered within search results.

This diagnosis helps avoid treating an SEO experiment as a generic publishing exercise. It also makes the eventual proposal easier for leadership to evaluate because it connects a defined problem to a contained action.

Put commercial intent ahead of raw traffic volume

A central theme in Berthold's advice is to focus on where revenue is generated. That can mean improving existing landing pages that target commercial or transactional queries, or publishing pages that address a clear market or topic gap.

As Search Engine Journal's interview with Berthold explains, he sees AI Overviews and AI Mode as increasingly taking top-of-funnel traffic. He also cautions that high-volume informational traffic does not necessarily contribute meaningfully to the bottom line.

This does not mean top-of-funnel content has no role. It can support brand awareness, entity understanding, internal linking, and later-stage decisions. But when resources are constrained, the initial test should usually prioritize the pages with a stronger connection to a commercial outcome.

Keyword tools can be useful here, but volume should not be the sole filter. Berthold's framework supports evaluating long-tail phrases and intent signals even when estimated volume is low. Search behavior is changing, and large language model-driven query expansion can surface needs that conventional keyword-volume estimates do not fully capture.

Build a test that leadership can approve

A low-budget SEO proposal is more likely to gain support when it is framed as a hypothesis with an explicit learning objective. Rather than saying, “We need to refresh these pages for more traffic,” define what the team expects to learn and why the answer will affect future work.

A practical hypothesis might follow this structure:

If we improve an existing commercial landing page to address a specific topic gap and search intent, we expect to see directional changes in impressions and relevant organic engagement. The result will help determine whether to extend the approach to similar pages.

This framing has several advantages. It limits perceived risk, identifies the decision that follows the test, and avoids presenting uncertain SEO outcomes as promises. It also clarifies that an unsuccessful result can still be useful if it rules out an assumption or reveals a technical, content, or intent mismatch.

Keep the initial scope narrow. A small group of comparable pages, one content format, or a defined topic gap is easier to monitor and less likely to create unintended sitewide effects. Only expand after the team sees repeatable evidence.

Use realistic evaluation windows

SEO does not provide the immediate feedback loop of paid media. Google must crawl updated pages before changes can influence rankings, impressions, and clicks. Berthold suggests using roughly 30 days to look for directional movement and up to 60 days to confirm whether a pattern is developing.

These are practical guidelines from an executive interview, not universal guarantees. Crawl frequency, competition, site architecture, seasonality, query demand, and the scale of the change can all affect results. Teams should record the date of each update and compare an appropriate pre-test baseline, while resisting the urge to declare success or failure too early.

Where AI can support the research process

AI can help reduce the research burden around a focused experiment. In the interview, Berthold points to AI-assisted topic discovery and the use of AI citation data to identify outreach opportunities and potential brand mentions.

For a commercial-content test, AI-assisted research may help teams organize questions customers ask, identify adjacent subtopics, or spot gaps between a page's current coverage and likely search intent. Citation data can also point to publications or sources that are already referenced in AI-generated answers, creating a starting point for relevant outreach.

The important distinction is that AI can accelerate discovery, not validate every conclusion. Teams still need to check search intent, verify claims, assess whether a proposed source is genuinely relevant, and ensure any new page gives users useful information beyond a reformatted version of existing content.

For AI search visibility, this work can complement traditional SEO. Clear, well-supported commercial content and relevant third-party mentions may improve the information available about a brand, but neither is proof of citation or recommendation in a particular AI system.

If flat traffic is making it difficult to decide where to invest next, GEO Search Leads can help you assess your website's GEO readiness and identify optimization opportunities around the content and signals that matter most. Use the findings to form a focused, evidence-led testing plan before expanding your effort. Create a GEO Search Leads account.

Frequently Asked Questions

What should an SEO team test first when organic traffic is flat?

Start with a site analysis to identify stalled areas and topic gaps. When budget is limited, prioritize existing or new pages that address commercial or transactional intent and have a clearer relationship to conversion.

Why focus on bottom-of-funnel SEO pages?

Berthold's framework argues that top-of-funnel traffic is increasingly affected by AI-driven search formats and may not always contribute directly to revenue. Bottom-of-funnel pages can offer a more practical starting point when the goal is to test commercially meaningful visibility.

How long should a low-budget SEO experiment run?

The interview suggests roughly 30 days to observe directional changes in impressions and up to 60 days to assess whether a pattern is holding. Actual timing can vary because Google must crawl changes and because search conditions differ by site and query.

Can AI citation data guarantee visibility in AI answers?

No. Citation data can help identify research and outreach opportunities, including possible sources of relevant brand mentions. It does not guarantee that a brand will be cited, recommended, or ranked by an AI search product.

Conclusion

When traffic is flat and resources are limited, the most defensible SEO response is a narrow, measurable experiment tied to commercial intent. Diagnose the gap, state the hypothesis, allow time for search systems to respond, and scale only when the evidence supports doing so. Berthold's guidance is a useful practitioner framework, though teams should adapt it to their own site architecture, market, and sales cycle.