Publishers who invested heavily in long-form content — 3,000 to 10,000-word comprehensive guides — reported a 25-45% traffic decline to their pillar content pages in verticals where AI Overviews expanded in 2024-2025. The irony is precise: the content that contains the most extractable information produces the most complete AI answers, which most effectively eliminates the need to visit the source. The pages with the highest content investment face the highest AI extraction risk. This creates an ROI inversion where content quality and content traffic value become inversely correlated for certain query types.
The extraction paradox: the most informative content produces the most comprehensive AI answers, generating the least click-through
AI systems synthesize the best available content into complete answers. When a long-form guide is the best source for a query, the AI system extracts enough to satisfy the user without requiring a click. Inferior content paradoxically faces less extraction risk because AI systems extract less from it, leaving more user motivation to click through for a complete answer. This extraction paradox creates a structural penalty for content quality in AI search.
The data confirms the paradox at scale. Ahrefs’ December 2025 update found that AI Overviews reduce position-one organic CTR by 58%, up from 34.5% in April 2025, with the impact worsening over time as AI extraction capabilities improve. Forbes experienced a 50% traffic decline year-over-year by July 2025. Learning platform Chegg reported a 49% decline in non-subscriber traffic as AI Overviews answered the educational queries that drove their traffic model. Wikipedia’s case illustrates the paradox most clearly: a 50% surge in bandwidth consumption from AI bots since January 2024, yet declining human engagement. The platform’s comprehensive, well-structured content makes it an ideal extraction target, which simultaneously makes it less necessary to visit.
The paradox operates most severely on informational queries with definitive answers. Queries asking “what is,” “how many,” or “what are the steps” receive comprehensive AI answers that reduce click motivation most effectively. Complex queries requiring nuanced judgment, multi-step processes with personalization requirements, or experiential evaluation remain partially extraction-resistant because the AI answer cannot fully substitute for the source content.
Long-form content ROI shifts from direct traffic value to indirect authority and citation value
When long-form content no longer drives proportional direct traffic, its value proposition shifts to indirect benefits: AI citation frequency building brand authority, passage extraction feeding entity representation in AI systems, and comprehensive topical coverage maintaining organic ranking for click-retaining query variants.
The ROI recalculation requires valuing indirect benefits alongside diminished direct traffic. Brands cited in AI Overviews earn 35% more organic clicks on their standard listings, meaning the long-form content’s citation presence lifts traffic to other pages even when the long-form page itself loses direct visits. AI search visitors who do click through convert at 14.2% compared to traditional organic’s 2.8%, and spend 68% more time on-site. The per-visit value of AI-referred traffic is substantially higher even though visit volume declines.
The content investment threshold where long-form content remains ROI-positive shifts depending on the indirect value capture. A 5,000-word guide that previously generated 10,000 monthly visits at $0.50 per visit equivalent may now generate 6,000 monthly visits but contribute to 200 AI citations producing brand impressions worth $2-5 each. The direct traffic ROI drops, but the combined direct-plus-indirect ROI may exceed the previous pure-traffic model. Content teams that only measure direct traffic will see a negative ROI trend and reduce investment, missing the indirect value their competitors will capture.
Content gating and progressive disclosure create click-motivation barriers that AI extraction cannot bypass
Structuring long-form content with publicly accessible summary sections and gated or interaction-requiring depth sections creates a natural click-motivation barrier. The AI system can extract the summary but cannot replicate the gated value. This preserves both AI citation presence (through the extractable summary) and direct traffic value (through the gated depth).
Progressive disclosure strategies that work without damaging organic ranking include embedding interactive calculators or assessment tools within the content that require page visits to use, providing summary data points publicly while gating the full dataset downloads, offering personalized analysis sections that require user input to generate results, and including case study details or implementation templates accessible only on the page. Search Engine Journal’s 2025 analysis confirms that publishers adapting to AI Overviews increasingly use this hybrid approach.
The balance between extractable and non-extractable sections is critical. Too much gating reduces the content’s extractability, eliminating AI citation potential. Too little gating leaves the full content available for extraction, eliminating click motivation. The effective ratio places approximately 30-40% of the content’s unique value in the publicly extractable layer (enough for meaningful AI citations) and 60-70% in the interaction-requiring layer (enough to motivate clicks for users who need deeper engagement).
The ROI floor: long-form content retains full value for queries where AI answers are suppressed or insufficient
Not all queries trigger AI answers, and not all AI answers are comprehensive enough to satisfy intent. Long-form content targeting complex, multi-faceted, or experiential queries retains pre-AI ROI levels because AI systems cannot adequately synthesize answers for these query types.
Semrush’s 2025 study tracking AI Overview prevalence across millions of keywords found that prevalence peaked at approximately 25% in July 2025 before settling at 15.69% by November. This means roughly 84% of queries still do not trigger AI Overviews, preserving traditional content ROI for the majority of the search landscape. Even among queries that trigger AI Overviews, the Dataslayer analysis found a counterintuitive result: zero-click rates did not universally increase for all AI Overview queries, challenging the assumption that every AI Overview destroys click-through.
The query characteristics that predict AI extraction resistance include multi-step processes requiring sequential implementation, subjective evaluations requiring personal judgment, localized or personalized information needs, rapidly changing topics where AI training data lags, and professional-grade technical procedures where incomplete answers carry risk. Shifting long-form content investment toward these extraction-resistant topic categories preserves the traditional ROI model while maintaining citation presence through shorter, data-rich content for extraction-vulnerable informational queries.
What is the extraction paradox and how does it affect high-quality content?
The most informative content produces the most comprehensive AI answers, which most effectively eliminates the need to visit the source page. Inferior content paradoxically faces less extraction risk because AI systems extract less from it, leaving more user motivation to click through. This creates a structural penalty for content quality in AI search, where the pages with the highest content investment face the highest traffic displacement from AI extraction.
What is the recommended ratio of extractable to gated content for preserving both AI citations and click-through?
The effective ratio places approximately 30-40% of the content’s unique value in the publicly extractable layer, enough for meaningful AI citations, and 60-70% in the interaction-requiring layer, enough to motivate clicks for users who need deeper engagement. Too much gating eliminates AI citation potential entirely, while too little gating leaves full content available for extraction, removing all click motivation from satisfied users.
What percentage of queries still do not trigger AI Overviews, preserving traditional long-form content ROI?
Semrush’s 2025 tracking found that AI Overview prevalence settled at approximately 15.69% by November 2025, meaning roughly 84% of queries still do not trigger AI Overviews. Long-form content targeting complex, multi-faceted, or experiential queries retains pre-AI ROI levels because AI systems cannot adequately synthesize answers for these query types. Shifting long-form investment toward extraction-resistant topic categories preserves traditional ROI.
Sources
- https://ahrefs.com/blog/ai-overviews-reduce-clicks-update/
- https://www.searchenginejournal.com/impact-of-ai-overviews-how-publishers-need-to-adapt/556843/
- https://thedigitalbloom.com/learn/2025-organic-traffic-crisis-analysis-report/
- https://discoveredlabs.com/blog/google-ai-overviews-traffic-impact-measuring-roi-pipeline-attribution
- https://www.semrush.com/blog/semrush-ai-overviews-study/
