AI Referral Traffic 2026: Data, Impact & Strategy

AI referral traffic describes website visits that originate directly from answers generated by large language models such as ChatGPT. This traffic channel grew by 206 % in 2025 and reached an all-time high in May 2026. Brands recommended by ChatGPT receive 2.5 times more visits than non-recommended competitors, according to a Similarweb study – yet the majority of those visits appear in analytics dashboards as organic branded search, not as referral. This article contextualises the current evidence from six independent studies, quantifies the impact of AI recommendations on real website traffic, and outlines the strategic implications for B2B brands.

Why AI referral traffic is now a commercial metric

Search engines are no longer the only place a brand gets found. The mechanics of visibility now stretch across Google's index and the answers that ChatGPT, Perplexity and comparable models generate – and treating those as two separate problems means solving neither. Crispy Content shows how agentic SEO and generative engine optimisation work as one data-driven, automated discipline, rather than as the next label to chase. The interesting question is not whether the term "GEO" will survive; the interesting question is how a brand earns a place in an AI-generated answer at all – and that mechanism is worth understanding before the next acronym arrives.

The evidence base has shifted from anecdotal to robust within a matter of months. Six studies – from Semrush to Similarweb to DACH-specific research – now deliver converging figures for the first time. Ignoring these numbers means making budget decisions on an incomplete picture. Overstating them means confusing growth rates with absolute volumes. Both pitfalls are avoidable once we lay the data side by side and name its limitations.

What is AI referral traffic – and why does it barely show up in Google Analytics?

AI referral traffic occurs when a user clicks a link inside an LLM answer and the destination site registers the referrer – such as chatgpt.com or claude.ai – in GA4. This measurable share, however, represents only a fraction of the actual impact. The larger attribution gap lies in what happens next.

Distinction from AI Overviews: Google's AI Overviews do not generate a separate referrer. A click from an AI Overview appears in GA4 as a regular organic visit. AI referral traffic in the strict sense originates exclusively from standalone LLM platforms with their own domain as referrer.

The Similarweb panel study reveals the real problem: 55.9 % of AI-influenced website visits do not appear in analytics as referral traffic but as branded search. The user reads a ChatGPT recommendation, remembers the brand name, opens a new tab and searches for the brand on Google. The last-click model attributes this visit to organic search – the upstream AI recommendation remains invisible.

Worked example: A B2B brand records 10,000 monthly branded searches in GA4. If the Similarweb factor of 55.9 % applies even approximately, up to 5,590 of those visits may have been triggered by an upstream AI recommendation – without any trace in the dashboard. The actual number depends on industry, brand awareness and product complexity. But the order of magnitude is clear: reading only the referral report means systematically underestimating the channel.

How fast is ChatGPT traffic actually growing? The numbers from three studies

Three independent studies using different methodologies arrive at converging results: ChatGPT referral traffic is growing at triple-digit rates per year, accelerated again in spring 2026, and reached an all-time high across all measured regions in May 2026. The studies differ in data sources and timeframes – which is precisely what makes their agreement meaningful.

Semrush clickstream analysis: 206 % growth in 12 months

Semrush and Datos analysed 1 billion clickstream data points from US users between October 2024 and February 2026. Referral traffic from ChatGPT to external websites grew year-on-year (January 2025 vs. January 2026) by 206 %. The number of domains receiving at least one ChatGPT referral rose from 71,000 to up to 260,000. At the same time, 30 % of all referrals concentrate on just 10 domains – a pattern reminiscent of the power distribution in organic search.

SE Ranking: all-time high in May 2026

The SE Ranking analysis covers 101,574 websites across 250 countries. The global AI referral share of total traffic rose by 36.7 % from April to May 2026 alone, reaching an all-time high. For the DACH region, the German-language study reports an increase from 0.23 % to 0.32 % year-on-year – growth of 39.1 %. In absolute terms, the share remains small. In growth dynamics, it is unmatched by any other channel.

Previsible: 9.9× growth in 19 months

Previsible analysed 6.77 million LLM sessions from 166 GA4 properties over 19 months. Monthly LLM sessions rose from 65,249 in November 2024 to 644,478 in May 2026 – a 9.9× increase. ChatGPT accounts for 92.4 % of all LLM referral sessions. Single-platform dominance at this level is unusual and carries strategic consequences for prioritisation.

What impact does an AI recommendation have on real website traffic?

An AI recommendation is not an awareness signal that may eventually convert. The Similarweb panel study documents a direct, measurable downstream effect: brands recommended by ChatGPT in its answers receive 2.5 times more website visits within seven days than comparable brands without a recommendation. The study ran from July 2025 to January 2026 based on a US desktop clickstream panel.

Engagement data reinforce the picture. AI-influenced visitors view 12 pages per visit and spend 11.8 minutes on site – compared with 6.5 pages and 5.6 minutes for regular visitors. These users arrive with a specific intent that the LLM has already pre-qualified. They are no longer searching – they are evaluating.

Limitation that must be stated: The Similarweb data come exclusively from US desktop users. Mobile usage – which accounts for a significant share of ChatGPT interactions – is not captured. Transferability to the DACH market is plausible but unproven. Anyone shifting budgets on the basis of these figures should use their own GA4 data as a cross-check.

Where does ChatGPT referral traffic go – and which industries benefit?

The distribution of ChatGPT referral traffic follows a power law: 30 % of all referrals land on just 10 domains, with Google alone absorbing 21.6 %. For the remaining 260,000 reachable domains, a fragmented but growing share remains. The differences between industries are substantial – and instructive for B2B brands.

The highest AI penetration occurs in industries with high information complexity and comparison needs. E-commerce grew by a factor of 37 since December 2024 – driven almost entirely by ChatGPT. A structural pattern emerges: 60 % of AI referral traffic lands on homepages, compared with just 17 % for organic search. This means LLMs link to brands, not to subpages. Any brand that fails to optimise its homepage as an entry point for pre-qualified visitors is leaving the channel on the table.

For B2B brands, the implication is clear: product pages, comparison pages and structured information architectures increase the likelihood that an LLM surfaces the brand as a recommendation. Producing blog content alone serves the model's information needs – but not its recommendation mechanism.

Why May 2026 marks a turning point

May 2026 is not an arbitrary data point. Three product decisions by OpenAI fell within the same period and produced a synchronous spike across all measured regions. This is not organic growth – it is a platform effect.

• New link format (7 May 2026): ChatGPT now displays brand names as directly clickable links within answers. The click path shortens from "read brand name → copy → search" to a single click. The conversion rate from reading to visiting rises mechanically.
• GPT-5.5 Instant as default model (5 May 2026): The new model generates structured recommendations with source citations more frequently. More answers contain links – more links generate more referrals.
• Self-serve ads manager: OpenAI opened an advertising system for self-service buyers in the same period. The commercial infrastructure signals that referral traffic will not remain a by-product but will be monetised.

The result: US +23.2 %, UK +38.7 %, EU +42.7 % month-on-month. For brands, this turning point means AI visibility is shifting from pure awareness to measurable acquisition. What was previously a "nice to have" in brand monitoring is becoming a plannable traffic source – with all the dependencies that entails.

Market shares of AI platforms: ChatGPT dominates, but the market is diversifying

ChatGPT holds 92.4 % of all LLM referral sessions according to Previsible – or 75.96 % according to SE Ranking, which includes a broader platform base. In both cases, the dominance is overwhelming. But the margins are moving, and movement at the margins is strategically relevant.

Claude recorded 64× growth and overtook Perplexity in March 2026. Gemini is growing steadily at 3.2×, without the volatility of smaller platforms. Perplexity – once positioned as "the search engine among LLMs" – lost 61 % since its peak in March 2025. Copilot, at −96 % since peak, is effectively irrelevant for referral traffic.

The strategic consequence: optimising for AI visibility means, de facto, optimising for ChatGPT. Diversifying towards Claude and Gemini makes sense as a hedge, but not as a primary lever. At the same time, the Perplexity decline shows that market shares in this space can shift within quarters. Betting an entire strategy on a single platform repeats the mistake of Google dependency – only faster.

What the data does not show – and why that matters

Converging growth figures from six studies constitute a strong signal. But they do not answer the question that matters commercially: Does this traffic convert? None of the available studies provide conversion data. We know that AI-influenced visitors stay longer and view more pages. We do not know whether they buy, enquire or sign up.

A robust causal analysis is equally absent. The Similarweb study shows correlation – recommended brands receive more visits. Whether the recommendation is the cause, or whether already well-known brands are recommended more frequently, cannot be separated from the data. Both effects likely operate simultaneously.

For the DACH region, an additional caveat applies: 0.32 % AI share of total traffic is a small number. For a website with 100,000 monthly visits, that amounts to 320 sessions. The growth is impressive; the volume is not – yet. Anyone diverting budget from proven channels to fund AI visibility today needs a longer time horizon than a single quarter.

The honest summary: this channel is growing faster than any other. It is real, measurable and directionally unambiguous. It is not yet large enough to replace existing channels – but it is large enough that ignoring it comes at a cost.

Strategic implications for B2B brands

The evidence supports three conclusions that follow from the data without speculation. First: measure AI referral traffic before attempting to optimise it. Any organisation that has not set up a separate channel in GA4 for chatgpt.com, claude.ai and gemini.google.com cannot see the channel – and cannot make an informed decision. Second: the attribution gap of 55.9 % means branded search must be re-evaluated as a KPI. A rise in branded searches with no identifiable campaign may be an AI effect. Third: the homepage is the primary entry point for LLM referrals. Structured, machine-readable information on the homepage – positioning, value proposition, differentiation – increases the probability of a recommendation.

Methods deliver guarantees. But only when the data foundation they rest on is sound. The six studies evaluated here provide that foundation – with stated limitations, converging trends and a clear direction. The rest is execution.

Sources

Semrush / Datos (2026): ChatGPT Traffic Analysis: Insights from 17 Months of Clickstream Data. URL: https://www.semrush.com/blog/chatgpt-search-insights/ (accessed 20 July 2026).

SE Ranking (2026): Referral Traffic from ChatGPT Hit an All-Time High in May 2026. URL: https://seranking.com/blog/chatgpt-referral-traffic-may-2026/ (accessed 20 July 2026).

Previsible (2026): 2026 State of AI Discovery Report: ChatGPT Wins 92% Share. URL: https://previsible.com/seo-strategy/ai-traffic-report-july-2026/ (accessed 20 July 2026).

Similarweb (2026): The Downstream Impact of AI Visibility (reported via ALM Corp). URL: https://almcorp.com/de/news/chatgpt-ai-recommendations-2-5x-traffic-similarweb-study/ (accessed 20 July 2026).

SE Ranking (2026): Study: AI Traffic in the DACH Region 2026. URL: https://seranking.com/de/blog/ai-traffic-dach-studie/ (accessed 20 July 2026).

Ahrefs (2026): Update: AI Overviews Reduce Clicks by 58%. URL: https://ahrefs.com/blog/ai-overviews-reduce-clicks-update/ (accessed 20 July 2026).