The decline in organic traffic is real. It has been measured, reported and felt across sectors. What is less clear - and far more consequential - is whether AI search is picking up the slack. The evidence, increasingly, suggests it is not.
That gap between traffic lost and traffic gained through AI surfaces is the central problem for any business running an AEO programme right now. Getting cited in Google AI Overviews, ChatGPT or Perplexity is genuinely valuable. But it does not automatically translate into the click volumes that traditional organic rankings once delivered. The mechanics are different. So is the measurement.
The Traffic Replacement Myth
When Google AI Overviews rolled out globally, the working assumption in many marketing teams was that AI visibility would compensate for organic losses. If your content was being cited, the reasoning went, you were still in the game. That assumption deserves serious scrutiny.
AI Overviews answer queries directly on the results page. Perplexity synthesises sources and presents a conclusion. ChatGPT responds conversationally without a list of links to click through. Each of these surfaces is designed to reduce the need for the user to leave the platform. That is not a flaw in the system - it is the point. But it means that a citation in an AI response and a top-10 organic ranking are not equivalent in traffic terms, even if they feel equivalent in visibility terms.
The implication is uncomfortable: organic traffic can fall, AI visibility can rise, and revenue attributed to search can still decline. These three things can be simultaneously true, and businesses that conflate AI visibility with AI-driven traffic will be slow to spot the problem.
Where the Traffic Actually Goes
Search behaviour has shifted in a way that makes simple before-and-after comparisons unreliable. Users who previously clicked through to a website to answer a factual question may now get that answer directly in an AI Overview. Those sessions never happen. They are not lost to a competitor - they are simply not generated. That is a structurally different problem to losing a ranking position.
For transactional and higher-intent queries, the picture is more nuanced. AI Mode, for instance, is increasingly surfacing product options, service comparisons and local recommendations in ways that do drive clicks - but often to specific pages rather than homepages, and through AI-mediated referral paths that standard GA4 configurations do not capture cleanly. The traffic exists. It is just not being attributed correctly.
This creates a measurement illusion. Teams see declining organic sessions and interpret it as a pure loss. They do not see the AI-referred sessions that are arriving under direct or referral labels, miscategorised because the UTM parameters were absent or because the referrer string from ChatGPT or Gemini was not configured as a recognised source. The actual picture is worse than teams realise on one dimension, and better than they realise on another.
What Measuring AI Traffic Actually Requires
Fixing the measurement problem is not glamorous work, but it is the prerequisite for everything else. The first step is ensuring that known AI referrer domains are captured in your analytics configuration. Traffic from ChatGPT arrives via chat.openai.com, traffic from Perplexity via perplexity.ai, and Gemini referrals appear under their own domain. None of these are identified as AI search by default in GA4.
Beyond session-level attribution, the more meaningful question is whether AI-referred visitors convert, and what they are worth relative to organic or paid search visitors. That requires tracking referral sources through to CRM entries and matched revenue - not just sessions or even form fills. A business running a proper AEO programme should be able to report, at least directionally, what AI search is contributing to pipeline. Without that, AEO is an activity, not a programme.
Google Search Console now surfaces some data on AI Overview impressions and clicks, which provides a partial view of AI-influenced performance within Google's ecosystem. That data is useful but incomplete - it tells you nothing about ChatGPT, Perplexity or Gemini referrals, which sit entirely outside GSC's scope. A full measurement picture requires combining GSC data with referral source analysis in GA4 and CRM pipeline matching.
The Visibility That Still Matters
The fact that AI citations do not always produce direct clicks does not make them worthless. AI search is increasingly where trust is built before a purchase decision is made. A user who sees your brand cited accurately in a Perplexity research summary, or recommended by ChatGPT in response to a category query, is more likely to search for you directly, engage with your paid ads, or convert when they eventually reach your site.
That assisted influence is real. It is also genuinely difficult to measure with the tools most businesses currently have in place. The practical implication is that AEO should be treated as a pipeline-influencing activity rather than a direct traffic channel - at least for now, and at least for informational query types. For transactional queries and local intent, AI-referred traffic is more direct, more attributable, and more worth optimising for.
Structuring content to be AI-extractable - clear factual claims, well-marked entities, schema where relevant, concise answers to specific questions - serves both goals simultaneously. It improves the likelihood of citation and, where the AI surface does drive clicks, it improves the quality of the landing experience. That is a reasonable place to anchor an AEO content strategy.
What to Do When Traffic Falls and AI Rises
If your organic traffic is falling and your AI visibility appears to be growing, the first priority is to confirm whether that visibility is real and consistent, or sporadic and query-specific. Manually querying your key topics in Google AI Overviews, ChatGPT, Perplexity and Gemini gives you a ground-level view that automated rank trackers do not yet reliably replicate across all platforms.
The second priority is to segment the organic traffic decline. Identify which query types have lost the most - informational, navigational or transactional. Informational query losses are often AI-driven and may not be recoverable through traditional optimisation. Transactional losses need a different diagnosis: check whether AI Overviews are featuring competitors on those terms, and whether your product or service pages are structured in a way that AI can extract and cite specific details.
Neither SEO nor AEO is finished. But the two disciplines are no longer measuring the same thing, and the assumption that growing AI visibility compensates for falling organic traffic needs to be tested with data rather than taken on faith. The gap between those two trends is where the real strategic work sits.