Google has started to share data on how user behaviour is changing inside AI Mode, and the signals are meaningful. According to Google's Ads Decoded series, queries in AI Mode are getting longer and more conversational, users are entering the research process earlier than they traditionally would have, and there is a growing appetite for human perspectives rather than generic informational content. Each of those shifts has a direct implication for how brands should structure their content if they want to earn citations.
This is not speculation about where AI search is heading. Google is describing what is already happening. For teams running AEO programmes, it reframes the question from 'how do we appear in AI answers?' to 'which content types are actually being surfaced for the queries AI Mode is now handling?'
The Query Length Shift Changes What Gets Cited
Shorter queries tend to surface short answers. When a user types a three-word search, AI Overviews can pull a definition or a quick comparison from almost anywhere. Longer, more complex queries require more nuanced responses, and the content that gets extracted tends to be more structured, more specific, and more authoritative.
If AI Mode users are now asking questions like 'what should I consider when switching accountancy software mid-financial year for a 50-person business', the content that earns a citation has to actually address that scenario. A generic 'benefits of accountancy software' page will not be extracted. A structured guide that works through the timing risks, data migration considerations, and staff training implications might be. The query length shift rewards specificity.
This has practical implications for content planning. Teams should be auditing their existing content against the longer, more qualified queries their target customers are likely to ask - not just the short-tail terms they historically optimised for. Tools like Google Search Console will not yet show you AI Mode query data in a clean form, but you can use conversational prompts in ChatGPT, Perplexity, and Gemini to surface what kind of content is being cited in response to those complex queries right now.
Earlier Discovery Means Earlier Funnel Content Matters More
Google's data points to users entering research earlier in the buying cycle through AI search. That is a significant shift from the traditional search model, where brands competed hard for mid and lower funnel queries and treated top-of-funnel as a brand exercise. If AI Mode is becoming the entry point for early research, brands that are only structuring content for decision-stage queries are invisible at the moment users are forming their consideration sets.
Early discovery content tends to be category-level and educational. It answers questions like 'how does X work', 'what are the trade-offs between X and Y', or 'what should someone in my situation think about before choosing'. This is content that many brands have historically underprioritised because it does not convert directly. In an AI search context, it builds the citations that shape how an AI system represents your brand when a user is still at the awareness stage.
The attribution challenge here is real. If a user first encounters your brand as a cited source in an AI Mode answer three weeks before they convert through a different channel, your CRM is unlikely to record that touchpoint. Teams running AEO measurement programmes need to account for this by tracking AI-referred sessions separately, tagging them through UTM parameters where possible, and looking at assisted conversion data to understand whether AI visibility is contributing to pipeline even when it is not the last click.
The Human Perspective Signal Is Not Soft
Google's data highlights growing demand for human perspectives in AI search results. This is consistent with what practitioners have observed across AI Overviews and AI Mode - first-person accounts, expert commentary, and genuine opinions are getting cited at a higher rate than purely informational or product-led content.
The mechanism makes sense. AI systems are trained to synthesise information, and they are reasonably good at presenting consensus views. What they cannot generate convincingly is an authentic human opinion grounded in direct experience. When a user asks an AI for a recommendation and the AI includes a cited source, that source is often one that provided a clear, attributed perspective rather than a neutral summary.
For brands, this means that author attribution, practitioner bylines, and content that takes a genuine position are not just editorial preferences - they are citation signals. If your content team is producing pieces that deliberately avoid strong viewpoints to stay neutral, they may be making themselves less extractable by AI systems looking for authoritative human perspectives. This is one area where the traditional content marketing instinct to hedge can actively work against AI visibility.
What This Means for Content Structure Right Now
The combination of longer queries, earlier discovery, and demand for human perspective points to a specific type of content performing well in AI Mode: in-depth, scenario-specific guides written from an expert point of view, structured so that individual sections can be extracted cleanly. That means clear H2 and H3 hierarchies, direct answers near the top of each section, and content that does not bury the answer in three paragraphs of preamble.
Schema markup still matters, but it does not compensate for weak content. Article schema with author information, FAQ schema where appropriate, and clean entity signals around your organisation and subject matter expertise all help AI crawlers understand the context of what they are reading. But the extraction decision is primarily driven by how well the content actually answers the query.
One practical exercise: take ten of the longest, most complex questions your sales team regularly hears from prospects and run them verbatim through AI Mode, ChatGPT, and Perplexity. Note which brands and sources are being cited. If you are not appearing in those responses, you have a clear content gap. The queries Google is describing as growing in AI Mode are likely very close to what your customers are already asking.
Planning for 2027 Without Waiting for Perfect Data
Google framing this research in the context of 2027 planning is useful. It signals that the query behaviour shifts are expected to continue and accelerate, not plateau. Teams that start building longer-form, scenario-specific, perspective-driven content now will have an established citation footprint by the time these patterns become dominant.
The measurement side needs to keep pace. AI-referred traffic from Google AI Mode does not always arrive with clean source attribution in GA4. Reviewing your referral sources regularly, setting up segments for known AI crawler patterns, and checking for dark traffic growth alongside your organic programme will give you a more accurate picture of whether your AEO investment is building traction. Revenue attribution is the end goal - tying AI-referred sessions through to CRM opportunities and closed deals is the only way to justify the programme commercially.
The underlying point from Google's own data is straightforward: AI Mode is changing what questions users ask and when they ask them. Brands that structure their content around the old query model will find themselves cited less often. Those that build for longer, more complex, earlier-stage queries - with genuine human expertise at the centre - are positioning themselves correctly for where AI search is heading.