We analyzed 15.4 million B2B SaaS sessions. Here’s what we found about AI referrals

Executive summary

AI search is becoming another way for B2B SaaS buyers to discover and evaluate software. But how much traffic do AI assistants actually send to company websites, and what happens when those visitors don’t convert immediately?

We analyzed 15.4 million website sessions across 150 B2B SaaS websites between October 2025 and September 2026, examining identifiable AI referral traffic, acquisition patterns, conversions, and recorded customer journeys.

AI referrals accounted for just 0.88% of total website sessions, despite being detected on 96% of the websites studied. ChatGPT was the largest identifiable source, contributing 65% of AI referral traffic.

The more interesting finding emerged when we looked beyond the initial visit.

Of 3,758 conversions associated with an identifiable AI referral, 36.4% occurred in a later session. Among those later conversions, 83.7% happened during visits attributed to direct traffic or organic search.

This highlights a measurement challenge for B2B SaaS marketers. When buyers discover a product through an AI assistant but return through another channel before converting, last-touch attribution may credit the returning channel rather than the earlier AI interaction.

The study also examines how AI referral traffic varies by company size, which AI platforms generate measurable visits, and where referral-based analytics cannot fully capture AI-assisted discovery.

Key findings

1. AI referrals accounted for 0.88% of website sessions. Across 15.38 million sessions, 136,080 were attributed to identifiable AI referral sources. AI traffic was detected on 144 of the 150 websites studied (96%).

2. ChatGPT generated 65% of identifiable AI referral traffic. Perplexity followed at 16.9%, while Gemini accounted for 9.1%.

3. More than one-third of AI-associated conversions occurred in later sessions. Of 3,758 conversions linked to an observed AI referral, 2,389 occurred during the initial session and 1,369 (36.4%) occurred later.

4. Direct and organic search accounted for 83.7% of later conversions. Among the 1,369 later conversions, 744 were recorded during direct visits and 402 during organic search visits. This shows how attribution credit can shift between channels across a recorded customer journey.

5. Identifiable AI referral traffic increased by 223% over the study period. Monthly AI referral sessions rose from 5,760 in October 2025 to 18,609 in September 2026. This represents growth in referral volume, not necessarily AI’s share of total website traffic.

1. Introduction

AI assistants have introduced another way for B2B software buyers to discover and evaluate products.

A buyer might ask ChatGPT to recommend a marketing analytics platform, compare alternatives in Perplexity, or use Gemini to research software for a specific business problem. Some of these interactions lead to visits to vendor websites.

But measuring that activity is more complicated than identifying another traffic source.

Website analytics can recognize visits when an AI platform passes identifiable referral information or tracking parameters. However, they cannot automatically capture every recommendation, conversation, or interaction that happened inside an AI assistant.

This matters because B2B buying journeys rarely end with a single visit.

A prospect might arrive through ChatGPT, explore a product’s features, leave without converting, and return through Google several days later to request a demo.

Depending on the attribution model, that conversion may be credited to organic search, the earlier AI referral, or both interactions.

To better understand these patterns, we analyzed 15.4 million website sessions across 150 B2B SaaS websites from October 2025 through September 2026.

The research focuses on three questions: how much identifiable traffic AI platforms generate, how those visitors convert during their initial and subsequent visits, and how returning through different channels affects attribution reporting.

We also examine what website analytics can reliably measure about AI-assisted discovery and where visibility remains limited.

The goal is to provide a clearer picture of AI referral traffic and its role in recorded B2B SaaS customer journeys, without confusing measurable interactions with the full extent of AI’s influence on buying decisions.

2. Research questions

We organized the analysis around five research questions:

RQ1. How widespread is identifiable AI referral traffic across B2B SaaS websites?

We measure the proportion of websites receiving detectable AI referrals and the percentage of total website sessions attributed to those sources.

RQ2. Which AI platforms generate the most identifiable website traffic?

We compare referral sessions from ChatGPT, Perplexity, Gemini, Claude, Copilot, and other recognized AI sources.

RQ3. How does AI referral conversion performance compare with other acquisition channels?

We examine same-session conversion rates across AI referrals, organic search, direct traffic, paid channels, and other sources using comparable measurement definitions. We distinguish these rates from conversions associated with AI referrals that occur in later sessions.

RQ4. What happens after visitors arrive through an identifiable AI referral?

We examine initial visits, same-session conversions, later observed conversions, and the acquisition channels through which visitors return before converting.

RQ5. What can referral and customer journey data reveal about AI-assisted discovery, and what remains unobservable?

We distinguish identifiable AI referrals and recorded customer journeys from AI-assisted interactions that cannot be established through the available analytics data.

3. Definitions

The following terms explain how AI referral traffic, conversions, and customer journeys are measured throughout this study.

• Identifiable AI referral: A website session attributed to a recognized AI platform through available referral information or tracking parameters.
• AI referral share: The percentage of total website sessions attributed to identifiable AI referral sources.
• Conversion: A recorded website action that meets a defined conversion goal, such as a product signup, demo request, or lead submission. Conversion goals may vary across the websites analyzed.
• Same-session conversion: A conversion recorded during the same website session in which the visitor arrived through an identifiable AI referral.
• Same-session conversion rate: The percentage of AI-referred sessions that recorded at least one qualifying conversion during that session. This measure is used when comparing AI referral performance with other acquisition channels.
• AI-associated conversion: A conversion linked to a recorded customer journey that includes an identifiable AI referral, either during the initial AI-referred session or in a later connected session.
• Later observed conversion: A conversion occurring in a subsequent session that can be connected to an earlier identifiable AI referral within the observation and attribution windows.
• Returning channel: The acquisition channel recorded when a visitor returns to the website and completes a conversion after an earlier identifiable AI referral.
• Unobservable AI influence: AI-assisted discovery or evaluation that cannot be identified through the available referral information and recorded customer journeys.

Interpreting AI referral attribution

An identifiable AI referral establishes that a website visit was attributed to a recognized AI source. It does not prove that the AI interaction caused a subsequent conversion.

Similarly, a conversion occurring after an AI-referred visit does not necessarily mean that AI was responsible for the outcome.

For this reason, the study distinguishes between same-session conversion performance and conversions associated with recorded AI referral journeys.

These measures answer different questions and should not be combined when comparing conversion rates across acquisition channels.

The absence of an identifiable AI referral also does not establish that AI played no role in a buyer’s discovery or evaluation process.

4. Research findings

Finding 1: AI referrals reached 96% of B2B SaaS websites but accounted for less than 1% of traffic

AI referrals were detected across nearly every website in the study, although they represented a relatively small share of overall traffic.

Of the 150 B2B SaaS websites analyzed, 144 received at least one identifiable AI referral during the 12-month observation period.

Together, these websites recorded 136,080 AI referral sessions out of 15,379,902 total website sessions, putting AI’s measurable traffic share at 0.88%.

That means fewer than nine out of every 1,000 website sessions originated from an identifiable AI referral source.

The difference between reach and traffic contribution is important. AI referrals appeared across 96% of the websites studied, but they still accounted for less than 1% of combined traffic.

How does AI referral traffic vary by company size?

We divided the 150 websites into three groups based on employee count, with 50 websites in each group.

Company size	Websites	Websites with AI referrals	Total sessions	AI referral sessions	AI traffic share
1–50 employees	50	48	1,857,439	20,766	1.12%
51–200 employees	50	48	4,847,172	42,148	0.87%
201+ employees	50	48	8,675,291	73,166	0.84%
Total	150	144	15,379,902	136,080	0.88%

Smaller B2B SaaS companies recorded the highest proportion of identifiable AI referral traffic at 1.12%, compared with 0.84% among companies with more than 200 employees.

However, larger companies received substantially more AI referral sessions in absolute terms.

Websites in the 201+ employee group recorded 73,166 AI referral sessions, compared with 20,766 among websites with 1–50 employees.

This highlights an important distinction: a channel can contribute a larger percentage of traffic to smaller websites while generating more visits for larger companies.

These differences are descriptive. The study does not establish whether company size itself explains the variation in AI referral share. Factors such as content visibility, brand recognition, audience behavior, and overall traffic composition may also contribute.

AI referral traffic more than tripled over 12 months

Identifiable AI referral sessions increased considerably during the observation period.

In October 2025, the websites recorded 5,760 AI referral sessions. By September 2026, that figure had reached 18,609.

This represents an increase of approximately 223% between the first and last months, or 3.23 times the initial monthly volume.

Month	AI referral sessions
October 2025	5,760
November 2025	6,488
December 2025	7,305
January 2026	8,310
February 2026	9,529
March 2026	10,351
April 2026	11,732
May 2026	12,522
June 2026	13,580
July 2026	15,254
August 2026	16,640
September 2026	18,609

The monthly figures show a consistent increase in identifiable AI referral volume throughout the study period.

However, an increase in referral sessions does not necessarily mean AI accounted for a growing percentage of website traffic.

Establishing that would require comparing monthly AI referrals with total website sessions over the same period.

Changes in referral identification, tracking coverage, or the websites included in monthly reporting could also affect the observed trend.

What this finding means for B2B SaaS marketers

AI referrals are worth tracking as a separate acquisition source, even when they contribute a relatively small percentage of overall website traffic.

Three measurements can help marketers evaluate their contribution:

• AI referral prevalence: Whether a website receives identifiable traffic from AI platforms.
• AI traffic share: The percentage of total website sessions originating from recognizable AI referral sources.
• AI referral growth: How identifiable AI traffic volume and its share of overall traffic change over time.

Together, these metrics provide a clearer picture of AI referral activity than traffic volume alone.

Measurement limitation: The findings cover identifiable AI referrals, not every visit potentially influenced by AI assistants. Interactions that do not produce recognizable referral information cannot be reliably classified as AI-driven traffic using website analytics alone.

Finding 2: ChatGPT accounted for 65% of identifiable AI referral traffic

Not all AI platforms contribute equally to measurable website traffic.

Visitors can arrive from ChatGPT, Perplexity, Gemini, Claude, Copilot, and other AI-powered services. However, the number of identifiable referrals varies considerably across these platforms.

Across the 150 B2B SaaS websites analyzed, ChatGPT accounted for 65% of all identifiable AI referral sessions, making it the largest recorded source.

Perplexity followed at 16.9%, while Gemini contributed 9.1%. The remaining 9% came from Claude, Copilot, and other identifiable AI sources.

These figures describe the distribution of measurable AI referral traffic within the study, rather than the overall popularity or influence of individual AI platforms.

Why platform-level attribution matters

Grouping all AI referrals into a single acquisition channel can provide a useful overview, but it may hide meaningful differences between individual platforms.

For example, a B2B SaaS company might receive most of its AI referral traffic from ChatGPT while attracting fewer visitors from Perplexity or Gemini.

Those visitors may land on different pages, explore different features, or convert at different rates.

Tracking AI referral sources individually helps marketers:

• Identify leading referral sources: Understand which AI platforms generate the most measurable website visits.
• Compare visitor behavior: Examine landing pages, engagement, and conversions across different AI referral sources.
• Monitor changes over time: Identify whether referral traffic is increasing or declining for individual platforms.
• Evaluate customer journeys: Understand how visitors from different AI platforms interact with the website before converting.

However, referral volume should not be interpreted as a direct measure of how often an AI assistant recommends a company.

An AI platform may mention a product without generating a website visit. A buyer may also discover a company through an AI-generated answer and later visit its website through Google or direct traffic.

Neither interaction would necessarily appear as an identifiable referral from the original AI platform.

What this finding means for B2B SaaS marketers

ChatGPT generated the majority of identifiable AI referral sessions in this study, but that does not mean other AI platforms should be ignored.

Marketers should evaluate individual platforms based on their measurable contribution to website traffic and customer journeys rather than referral volume alone.

Separating AI sources also makes it easier to investigate changes in traffic quality, landing-page performance, and conversion behavior as AI-assisted discovery evolves.

Measurement limitation: AI referral identification depends on the source information available to the analytics system. Differences in referral behavior, browser environments, tracking parameters, and source classification can affect how much traffic is attributed to each platform. Visits without reliable source information should not be assigned to an AI platform without supporting evidence.

Finding 3: AI referrals recorded a 1.76% same-session conversion ratio

AI referrals accounted for less than 1% of website traffic across the 150 B2B SaaS websites studied. But traffic volume alone doesn’t tell us how visitors behave after arriving.

Across 136,080 identifiable AI referral sessions, the study recorded 2,389 conversions during the initial AI-referred session.

That represents approximately 1.76 conversions per 100 AI referral sessions.

The distinction between same-session and later conversions matters. An additional 1,369 conversions were associated with recorded journeys that included an earlier AI referral, but occurred during subsequent sessions.

Those later conversions are important for customer journey analysis. However, they should not be included in the numerator when calculating a same-session conversion rate.

How does AI referral performance compare with other acquisition channels?

To understand AI referral performance in context, we examined the reported conversion rates for other acquisition channels.

Acquisition channel	Reported conversion rate
AI referrals (same-session ratio)	1.76%
Organic search	2.02%
Direct traffic	1.93%
Paid channels	1.71%
Other channels	1.72%

These figures suggest that AI referral performance was broadly within the range recorded for other acquisition channels.

However, the comparison should be interpreted cautiously. The AI figure uses same-session conversions, while the other channel rates require confirmation that they use equivalent conversion definitions and counting methods.

Until those definitions are verified, the figures should not be used to establish a definitive ranking of acquisition channels.

Why conversion rates alone don’t tell the full story

A channel generating fewer visitors can still produce meaningful conversions. Similarly, a channel with a relatively high conversion rate may contribute fewer total conversions than one generating substantially more traffic.

For B2B SaaS companies, the type of conversion also matters.

A newsletter signup, product registration, qualified demo request, and sales opportunity represent different levels of commercial intent.

Comparing these actions as equivalent outcomes can create misleading differences between channels and companies.

This is particularly important when analyzing multiple B2B SaaS websites with different business models, audiences, and conversion goals.

A more complete evaluation should consider conversion volume, conversion type, and downstream business outcomes alongside conversion rates.

What about conversions occurring after the initial AI referral?

Same-session conversion reporting captures only part of the recorded customer journey.

In this study, 1,369 additional conversions were associated with earlier identifiable AI referrals but occurred during later sessions.

These conversions should be analyzed separately from same-session performance.

Their presence does not justify adding them to the initial AI session conversion rate. It does, however, raise an important attribution question: how should recorded interactions receive credit when visitors return through different channels before converting?

We examine this in Finding 4.

What this finding means for B2B SaaS marketers

AI referral performance should be evaluated using consistent measurement definitions rather than a single headline conversion rate.

Marketing teams should:

• Compare equivalent conversion events: Avoid treating product signups, demo requests, and other conversion goals as interchangeable outcomes.
• Separate session and journey reporting: Measure immediate conversions independently from conversions associated with earlier referral interactions.
• Examine company-level performance: Review whether aggregate results are influenced disproportionately by high-traffic websites.
• Connect conversions to business outcomes: Evaluate qualified leads, pipeline, and revenue where reliable downstream data is available.

The objective is to understand what identifiable AI referrals contribute without overstating their performance or overlooking their earlier presence in recorded customer journeys.

Measurement limitation: The 1.76% figure represents recorded same-session conversions divided by AI referral sessions. It is a true session conversion rate only if the numerator counts distinct converting sessions. Differences in conversion goals, website traffic composition, and counting methods may affect comparisons across acquisition channels.

Finding 4: Nearly 36% of AI-associated conversions happened after the initial visit

Not every visitor arriving from an AI assistant converts during their first session.

Some explore a product, leave the website, and return later through Google, direct traffic, or another acquisition channel before completing a conversion.

This creates an important attribution question: What happens to AI’s reported contribution when a visitor returns through another channel before converting?

Across the 150 B2B SaaS websites analyzed, the study identified 3,758 conversions associated with recorded customer journeys that included an identifiable AI referral.

Of those conversions:

• 2,389 (63.6%) occurred during the initial AI-referred session.
• 1,369 (36.4%) occurred during a later connected session.

This means more than one-third of the conversions associated with recorded AI referral journeys happened after the initial visit.

The finding highlights why measuring only immediate conversions may provide an incomplete picture of how identifiable AI referrals appear in customer journeys.

Which channels brought visitors back before converting?

We examined the acquisition channels recorded for the 1,369 conversions that occurred after an earlier identifiable AI referral.

Direct traffic accounted for the largest share, followed by organic search.

Returning channel	Later conversions	Share of later conversions
Direct	744	54.3%
Organic search	402	29.4%
Other recorded channels	223	16.3%
Total	1,369	100%

Together, direct and organic search accounted for 83.7% of later conversions associated with an earlier identifiable AI referral.

This does not mean those channels replaced AI’s role in the customer journey. It means they were the recorded acquisition sources for the sessions in which those conversions occurred.

How attribution models can change the reported contribution of AI referrals

Consider a visitor who arrives through ChatGPT, explores a product’s features, and leaves without converting.

Three days later, the same visitor searches for the company on Google, returns to the website, and requests a demo.

Both interactions are part of the recorded journey, assuming the analytics system can reliably connect the sessions.

However, different attribution models may assign conversion credit differently:

• First-touch attribution: Credits ChatGPT if the AI referral was the first eligible recorded interaction.
• Last-touch attribution: Credits organic search if it was the final eligible interaction before conversion.
• Linear attribution: Distributes credit equally across eligible recorded interactions.
• Time-decay attribution: Assigns more credit to interactions occurring closer to the conversion.

These models provide different perspectives on the same recorded journey.

None independently establishes which interaction caused the conversion.

Why session-level reporting can miss earlier AI interactions

The 1,369 later conversions demonstrate an important distinction between acquisition reporting and customer journey attribution.

A session-level report identifies the acquisition channel associated with the session in which a conversion occurred.

A journey-level attribution report can also account for eligible interactions recorded earlier in the visitor’s journey.

As a result, a conversion completed during an organic search or direct session may also be associated with an earlier identifiable AI referral.

However, assigning every later conversion entirely to AI would overstate what the data establishes.

The more useful approach is to examine how conversion credit changes across attribution models while keeping the underlying journey data consistent.

What this finding means for B2B SaaS marketers

Evaluating AI referrals only through immediate conversions can overlook their presence earlier in recorded customer journeys.

Marketing teams should investigate three questions:

1. How many conversions occur during the initial AI-referred session?
2. How many occur in later sessions connected to an earlier AI referral?
3. How does the reported contribution of AI change when different attribution models are applied?

These comparisons can help marketers understand the difference between immediate acquisition performance and the broader role of AI referrals in observable customer journeys.

They also reinforce why attribution analysis should consider multiple recorded interactions rather than relying exclusively on the channel associated with the converting session.

Measurement limitation: Later conversions can only be associated with earlier AI referrals when sessions are reliably connected. Consent settings, identity resolution, device changes, and attribution lookback windows may affect which journeys can be observed. The findings describe recorded associations, not the causal contribution of AI to conversions.

Finding 5: AI search’s measurable contribution doesn’t tell the full attribution story

The previous findings show that identifiable AI referrals generate website visits and appear in recorded customer journeys that lead to conversions.

But there’s an important limitation: not every interaction with an AI assistant produces a detectable website referral.

A buyer might ask ChatGPT to compare B2B SaaS platforms, remember a recommended product, and visit its website directly the next day.

Another buyer might research vendors through Perplexity before searching for a specific company on Google.

In both cases, AI may have contributed to product discovery. However, the website’s analytics data may contain no identifiable evidence of those earlier interactions.

This creates a gap between the AI activity marketers can measure and the broader influence they may want to understand.

What the study can measure

Across the 150 B2B SaaS websites analyzed, identifiable AI referrals accounted for 136,080 sessions, representing 0.88% of total website traffic.

The study also identified 3,758 conversions associated with recorded journeys containing an AI referral, including 1,369 conversions occurring in later connected sessions.

These findings provide evidence of measurable AI referral activity and its presence in recorded conversion journeys.

However, they do not establish how many additional buyers interacted with AI assistants before arriving through channels without identifiable AI referral information.

Three ways AI-assisted discovery can be difficult to attribute

1. AI recommendations that don’t generate a website click

A buyer can discover a software company through an AI-generated answer without immediately visiting its website.

If they later type the website address into their browser, the original AI interaction may not appear in the analytics data.

The website can record the direct visit, but it cannot reliably identify the earlier AI recommendation from that visit alone.

2. AI-referred visitors who return through another channel

Some AI-referred visitors leave without converting and return through another acquisition channel.

As Finding 4 shows, 36.4% of AI-associated conversions in the study occurred during later connected sessions.

Among those later conversions, direct and organic search accounted for 83.7%.

These interactions can remain visible in customer journey reporting when the analytics system can connect sessions reliably.

However, a report focused only on the converting session may not show the earlier AI referral.

3. AI traffic without identifiable source information

Not every visit originating from an AI-powered experience includes recognizable referral information.

Source identification can be affected by browser settings, application environments, redirects, tracking parameters, and analytics configuration.

When reliable source information is missing, a visit may appear under another acquisition category or remain unattributed.

That does not establish that the visit originated from AI. It means the available data is insufficient to determine its source.

What analytics can and cannot establish

The distinction between measurable AI referrals and broader AI influence is important when interpreting acquisition and attribution reports.

Measurement category	What analytics can establish	What remains uncertain
Identifiable AI referrals	Sessions attributed to recognized AI sources	Earlier interactions inside AI platforms
Same-session conversions	Conversions recorded during AI-referred sessions	Whether AI caused the conversion
Later observed conversions	Conversions connected to earlier AI referrals	The causal contribution of each interaction
Direct or organic visits without recorded AI touchpoints	The recorded acquisition source	Whether AI influenced earlier discovery
AI recommendations without website visits	Not measurable through website analytics alone	Recommendation exposure and subsequent influence

Can we estimate how much AI influence goes unmeasured?

Not from website referral and customer journey data alone.

The study measures identifiable AI referral sessions and conversions associated with recorded AI interactions.

It does not contain an independent record of every AI-generated recommendation, buyer prompt, or software evaluation that occurred before a website visit.

As a result, we cannot reliably estimate the percentage of buyers who used AI during their research but arrived without a detectable AI referral.

Answering that question would require additional evidence, such as buyer surveys, consented research panels, or other independently collected information about software discovery behavior.

Even then, self-reported information would need to be interpreted carefully.

For this reason, the 0.88% AI referral share represents identifiable AI traffic within the study, not AI’s total influence on B2B SaaS product discovery.

What this finding means for B2B SaaS marketers

The challenge isn’t simply identifying more AI referral traffic. It’s understanding what the available data can reliably establish.

Marketing teams can improve their measurement by:

• Separating identifiable AI referrals: Track recognizable AI sources rather than grouping them into general referral traffic.
• Connecting recorded customer journeys: Examine earlier interactions when consent and cross-session identification allow.
• Comparing attribution models: Understand how recorded AI interactions receive credit alongside other acquisition channels.
• Evaluating downstream outcomes: Connect identifiable referrals to qualified leads, pipeline, and revenue where reliable data is available.
• Acknowledging measurement gaps: Avoid assigning unidentified traffic to AI without supporting evidence.

These practices cannot reveal every AI-assisted interaction, but they can provide a clearer picture of the referrals and customer journeys that are actually recorded.

Measurement limitation: The study cannot quantify AI recommendations or buyer research activity that leaves no identifiable evidence in website analytics. The measurement gaps discussed here are limitations of referral-based attribution, not estimates of how much AI influence goes untracked.

5. Implications for B2B SaaS marketing teams

The findings highlight an important challenge for B2B SaaS marketers: AI referrals are becoming a measurable part of website acquisition, but understanding their contribution requires looking beyond traffic volume and immediate conversions.

Across the 150 websites studied, identifiable AI referrals accounted for just 0.88% of total sessions. Yet more than one-third of conversions associated with recorded AI referral journeys occurred during later visits.

These results point to five practical considerations.

5.1 Track AI referrals as a separate acquisition source

AI referrals should be identifiable in acquisition reports rather than grouped entirely under general referral traffic.

Separating traffic from ChatGPT, Perplexity, Gemini, and other AI platforms makes it easier to understand which sources generate visits and how their contribution changes over time.

However, referral volume should always be evaluated alongside total website traffic.

A growing number of AI referral sessions doesn’t necessarily mean AI is becoming a larger share of acquisition.

Marketers should monitor both absolute referral volume and the percentage of overall traffic attributed to AI sources.

5.2 Look beyond the initial website visit

One of the study’s most relevant findings is that 36.4% of AI-associated conversions occurred during later connected sessions.

Among those conversions, direct and organic search accounted for 83.7% of the recorded returning channels.

This highlights why session-level acquisition reports and customer journey attribution reports can tell different stories.

A conversion completed during an organic search visit may also have an earlier identifiable AI referral in its recorded journey.

Where reliable cross-session identification is available, marketers should compare first-touch, last-touch, and multi-touch attribution to understand how conversion credit changes across recorded interactions.

The objective isn’t to assign more credit to AI automatically. It’s to avoid overlooking earlier touchpoints when evaluating acquisition performance.

5.3 Compare conversion performance using consistent definitions

Conversion rates are only meaningful when the underlying measurements are comparable.

A product signup, demo request, and qualified lead represent different outcomes. Combining them without considering those differences can distort channel comparisons.

Marketing teams should define their conversion goals clearly and evaluate equivalent outcomes across acquisition sources.

They should also separate immediate conversions from those occurring later in the customer journey.

For B2B SaaS companies, this means looking beyond website conversion rates to understand whether visitors become qualified leads, enter the sales pipeline, or generate revenue.

5.4 Recognize the limits of AI referral attribution

Not every interaction with an AI assistant produces a detectable referral.

A buyer may discover a company through ChatGPT and visit its website directly several days later. Without an identifiable earlier interaction, website analytics cannot reliably establish that AI contributed to the visit.

This makes it important to distinguish measurable AI referrals from broader AI-assisted discovery.

Marketers should avoid assigning direct or unidentified traffic to AI based on assumptions.

Instead, they should report what can be observed and clearly acknowledge where attribution visibility ends.

5.5 Establish a consistent AI referral measurement baseline

The study recorded a 223% increase in monthly identifiable AI referral sessions between October 2025 and September 2026.

Although this reflects growth in referral volume rather than necessarily growth in traffic share, it reinforces the value of consistent measurement over time.

B2B SaaS marketing teams should establish a reporting baseline that includes:

• Identifiable AI referral sessions and their share of total traffic.
• Referral traffic by individual AI platform.
• Landing pages receiving AI-referred visitors.
• Same-session conversion performance by defined conversion goal.
• Later conversions associated with recorded AI referral journeys.
• Downstream qualified leads, pipeline, and revenue where available.

Reviewing these metrics consistently can help teams identify changes in AI referral activity and evaluate its contribution alongside other acquisition channels.

What the findings mean overall

AI search shouldn’t be evaluated solely by how many visitors it sends to a website.

The more useful question is how identifiable AI referrals appear throughout recorded customer journeys and how those interactions relate to meaningful business outcomes.

The study shows that a substantial share of AI-associated conversions occurred after the initial visit, often during sessions attributed to direct traffic or organic search.

For B2B SaaS marketers, this makes customer journey analysis an important complement to traditional acquisition reporting.

The goal is not to give AI more credit than the evidence supports, but to understand its measurable contribution without overlooking earlier interactions or confusing attribution with causation.

6. Methodology

This study examines identifiable AI referral traffic, conversion performance, and recorded customer journeys across 150 B2B SaaS websites.

The analysis covers 15,379,902 website sessions between October 1, 2025, and September 30, 2026.

We used descriptive analysis to examine AI referral prevalence, platform-level traffic distribution, conversion activity, and the acquisition channels associated with later conversions.

6.1 Sample selection

The study includes 150 B2B SaaS websites, organized into three groups based on company employee count.

Company size	Websites	Share of sample
1–50 employees	50	33.3%
51–200 employees	50	33.3%
201+ employees	50	33.3%
Total	150	100%

This grouping allows us to compare AI referral activity across websites of different sizes.

However, the sample should not be interpreted as a statistically representative cross-section of the entire B2B SaaS industry.

Company size, geography, business model, and analytics coverage may differ from those of the wider market.

6.2 Observation period and data coverage

The analysis covers 12 consecutive months, from October 2025 through September 2026.

The reported dataset includes:

• 15,379,902 website sessions.
• 136,080 identifiable AI referral sessions.
• 150 B2B SaaS websites.
• Three company-size segments.
• Monthly AI referral observations.
• Conversion outcomes and aggregated customer journey observations.

The 12-month observation window allows us to examine monthly referral patterns rather than relying on a single reporting period.

6.3 AI referral identification

An identifiable AI referral is a website session attributed to a recognized AI platform through available referral or campaign-source information.

The study includes the following AI referral categories:

• ChatGPT
• Perplexity
• Gemini
• Claude
• Microsoft Copilot
• Other identifiable AI sources

Sessions are classified as AI referrals only when the available source information supports that classification.

Visits without identifiable AI referral information cannot reliably be attributed to AI platforms based on assumptions about visitor behavior.

The resulting figures measure detectable AI referrals, not every interaction potentially influenced by AI-assisted discovery.

6.4 Traffic share and platform distribution

AI referral traffic share is calculated as:

AI referral share (%) = (Identifiable AI referral sessions ÷ Total website sessions) × 100

Across the study:

(136,080 ÷ 15,379,902) × 100 = 0.88%

Platform-level referral shares are calculated by dividing sessions attributed to each recognized AI source by the total number of identifiable AI referral sessions.

Company-size comparisons use the same traffic-share calculation within each employee-count group.

6.5 Conversion measurement

The analysis distinguishes between conversions occurring during an identifiable AI referral session and conversions occurring during later sessions associated with an earlier AI referral.

Same-session conversion ratio

The same-session conversion ratio is calculated as:

Same-session conversion ratio (%) = (Same-session conversions ÷ AI referral sessions) × 100

The study recorded 2,389 same-session conversions across 136,080 identifiable AI referral sessions.

(2,389 ÷ 136,080) × 100 = 1.76%

This represents recorded conversions per AI referral session. It should only be interpreted as the percentage of sessions that converted if the numerator represents distinct converting sessions.

Later observed conversions

Later conversions are reported separately from same-session conversion performance.

The study identified 1,369 conversions occurring during subsequent sessions associated with earlier identifiable AI referrals.

These conversions are included in customer journey analysis but are not added to the numerator when calculating the same-session conversion ratio.

Cross-channel comparisons

Conversion performance across acquisition channels should be compared using equivalent conversion definitions, counting methods, and observation periods.

Where conversion goals differ across websites, pooled conversion rates may reflect differences in business models or conversion-event definitions rather than acquisition-channel performance alone.

6.6 Customer journey analysis

The customer journey analysis examines conversions associated with recorded journeys containing an identifiable AI referral.

The reported conversion cohort contains 3,758 conversions, divided into:

• 2,389 conversions occurring during the initial AI-referred session.
• 1,369 conversions occurring during later connected sessions.

The share of later conversions is calculated as:

Later-conversion share (%) = (Later conversions ÷ Total AI-associated conversions) × 100

(1,369 ÷ 3,758) × 100 = 36.4%

The analysis also examines the acquisition channels recorded for later conversion sessions.

Of the 1,369 later conversions:

• 744 occurred during direct sessions.
• 402 occurred during organic search sessions.
• 223 occurred during sessions attributed to other recorded channels.

Together, direct and organic search accounted for 83.7% of later conversions.

These figures describe recorded associations between earlier AI referrals and subsequent conversions. They do not establish that AI caused those outcomes.

6.7 Monthly trend analysis

Monthly identifiable AI referral sessions were examined across the 12-month observation period.

The change between October 2025 and September 2026 is calculated as:

Referral growth (%) = [(Final-month sessions ÷ Initial-month sessions) − 1] × 100

Using the reported figures:

[(18,609 ÷ 5,760) − 1] × 100 = 223.1%

This represents the change in monthly identifiable AI referral volume between the first and last months.

It does not establish an equivalent increase in AI’s share of overall website traffic, which would require monthly total-session denominators.

6.8 Analytical approach

The study uses descriptive statistics to summarize acquisition and conversion patterns.

Research area	Measures
AI referral prevalence	Websites with identifiable AI referrals and percentage of sampled websites
Traffic contribution	AI referral sessions and share of total sessions
Platform distribution	AI referral sessions and percentage share by platform
Company-size comparisons	Sessions and AI referral share by employee-count group
Conversion performance	Same-session conversion ratios
Customer journeys	Same-session conversions, later conversions, and returning-channel distribution
Monthly trends	AI referral volume and first-to-last-month percentage change

The findings describe the recorded sample.

They should not be interpreted as causal estimates or statistically representative benchmarks for the wider B2B SaaS industry.

6.9 Data quality and interpretation

The analysis relies on accurate session counts, acquisition-source classification, conversion records, and the ability to associate later sessions with earlier AI referrals.

Several factors can affect these measurements, including missing referral information, inconsistent conversion goals, tracking interruptions, and limitations in cross-session identification.

Aggregate conversion comparisons may also be influenced disproportionately by websites with higher traffic volumes.

For these reasons, the reported findings are interpreted as descriptive associations within the analyzed sample.

6.10 Privacy and data handling

The study reports aggregated website acquisition and conversion findings.

Individual visitors and customer organizations are not identified in the published results.

The public analysis focuses on combined session counts, referral-source distributions, company-size groups, and aggregated conversion observations rather than individual customer records.

The availability of supporting data is addressed separately in Section 8.

7. Study limitations

Several limitations should be considered when interpreting the findings.

• Referral visibility: The study captures identifiable AI referrals, not every interaction involving AI-assisted discovery. Recommendations that do not generate detectable website visits remain outside the analysis.
• Sample representativeness: The analysis covers 150 B2B SaaS websites across three company-size groups. The findings should not be assumed to represent the entire B2B SaaS industry.
• Conversion definitions: Conversion goals may differ across websites. Product signups, demo requests, and lead submissions do not necessarily represent equivalent business outcomes, which can affect aggregate comparisons.
• Cross-session tracking: Later conversions can only be connected to earlier AI referrals when sufficient tracking and identity information is available. Consent settings, browser restrictions, device changes, and attribution windows may limit journey visibility.
• Aggregate reporting: Pooled results may be influenced disproportionately by high-traffic websites. They do not necessarily reflect the conversion performance of a typical B2B SaaS company.
• Attribution versus causation: An AI referral appearing earlier in a recorded customer journey does not establish that it caused the conversion. Attribution models assign credit to recorded interactions but cannot independently measure their incremental impact.
• Unobserved AI influence: The study cannot quantify AI-assisted research or recommendations that leave no identifiable evidence in website analytics.

8. Data availability

The underlying dataset used in this study is not publicly available due to customer confidentiality, privacy, and data protection considerations.

The research presents aggregated findings across the analyzed websites without identifying individual customers or visitors. Raw session records, customer-level data, and individual website datasets are not available for external access or independent analysis.

9. References

1. Semrush (2026). ChatGPT traffic analysis: Insights from 17 months of clickstream data. https://www.semrush.com/blog/chatgpt-search-insights/
2. Microsoft Clarity (2025). AI Traffic Converts at 3x the Rate of Other Channels (Study). https://clarity.microsoft.com/blog/ai-traffic-converts-at-3x-the-rate-of-other-channels-study/
3. Pew Research Center (2025). Do people click on links in Google AI summaries? https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/
4. Pew Research Center (2025). What Web Browsing Data Tells Us About How AI Appears Online. https://www.pewresearch.org/data-labs/2025/05/23/what-web-browsing-data-tells-us-about-how-ai-appears-online/
5. McKinsey & Company (2024). Five fundamental truths: How B2B winners keep growing. https://www.mckinsey.com/capabilities/growth-marketing-and-sales/our-insights/five-fundamental-truths-how-b2b-winners-keep-growing
6. Google Analytics. Campaigns and traffic sources. https://support.google.com/analytics/answer/11242841
7. Google Analytics. Scopes of traffic-source dimensions. https://support.google.com/analytics/answer/11080067
8. Google Analytics. Get started with attribution. https://support.google.com/analytics/answer/10596866
9. Google Analytics. About traffic-source dimensions. https://support.google.com/analytics/answer/15612152
10. Google for Developers. Google Analytics Data API: Dimensions and metrics. https://developers.google.com/analytics/devguides/reporting/data/v1/api-schema