AI Search Is Splitting in Two: Answer Engines and Assistants

A quiet split has become obvious in the last year: the tools people loosely call “AI search” have diverged into two products with different behavior, different citations and different optimization playbooks. Brands that measure them as one channel are drawing wrong conclusions from their own data.

The two surfaces

Answer engines — Google AI Overviews, Perplexity and similar tools — behave like a research layer on top of search. The user asks a question, the engine retrieves documents, and the answer arrives with visible citations. The citation list is the new page one, and being absent from it is the new page two.

Assistants — ChatGPT, Gemini, Claude, Copilot in their conversational modes — behave like advisors. They lean more on training knowledge and less on live retrieval, citations are often absent or loose, and recommendations emerge as shortlists (“what are the best tools for X?”). There is no citation list to fight for; there is a shortlist to be on.

The stakes, in numbers

The split is happening while the click economy underneath search shrinks. Pew Research Center’s March 2025 dataset — 68,879 Google searches by 900 U.S. adults — found 18% of searches produced an AI summary; on those pages, users clicked a traditional result 8% of the time versus 15% on pages without one, and clicked a source inside the summary itself just 1% of the time [1]. Ahrefs’ 300,000-keyword analysis puts the position-one click-through loss from AI Overviews at 34.5% in April 2025, widening to roughly 58% by December 2025 [2]. Meanwhile, Ahrefs’ study of 75,000 brands found branded web mentions correlate 0.664 with AI Overview presence — a stronger association than backlinks [3]. Attention is moving to the answer layer on both surfaces at once, which is exactly why the two surfaces have to be measured separately rather than averaged.

Why the split matters

• Different failure modes. On answer engines, a brand can fix visibility by fixing specific pages — extractable claims, entity clarity, corroborated facts. On assistants, visibility depends more on how widely and consistently the brand is described across the corpus the model learned from.
• Different timing. Page fixes can move answer-engine citations in days to weeks. Assistant behavior moves slower, because it tracks the broader web’s description of your brand, not one page.
• Different measurement. Answer-engine visibility is auditable per prompt: which sources were cited. Assistant visibility needs the same prompt run repeatedly, because shortlists vary between sessions and as models update.

What we see in monitoring data

Running the same buyer-shaped prompt set across both surfaces makes the split measurable. Three patterns come up repeatedly:

1. Brands cited consistently in AI Overviews are frequently absent from assistant shortlists for the same question — and vice versa. The two scores are correlated, not identical.
2. Assistant shortlists favor brands with a clear category label repeated across many independent sources. “What it is” has to be unambiguous before “how good it is” matters.
3. Position still matters on both surfaces. A mention in the first sentence of an answer is worth far more than a mention in the last paragraph, and this is true whether or not citations are shown.

The practical takeaway

Stop reporting “AI visibility” as a single number. Track answer-engine citations and assistant mentions as two separate trends, with their own prompt sets and their own improvement backlogs. The content work overlaps — clear claims and consistent entity descriptions help both — but the measurement cannot be merged, because the surfaces fail for different reasons.

If you want to see how your brand splits across the two surfaces today, Novaxilabs’ free baseline scan reports them separately, per engine, with the exact prompts used.

References

1. Pew Research Center, “Google users are less likely to click on links when an AI summary appears in the results,” July 2025. pewresearch.org
2. Ahrefs, “AI Overviews Reduce Clicks by 34.5%” (April 2025) and December 2025 update. ahrefs.com/blog/ai-overviews-reduce-clicks
3. Ahrefs, “An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied),” 2025. ahrefs.com/blog