The DOI matches its DataCite record.
AI Assistant Referral Traffic and Conversion Performance Relative to Organic Search: A 12-Month Multi-Site Benchmark, 2025-2026
Open accessAbstract
A 12-month benchmark (September 2025 to August 2026) of referral traffic from AI assistants (ChatGPT, Claude, Gemini, Perplexity, Copilot and similar) compared with organic search, using Google Analytics 4 data from 95 websites (66 meeting volume thresholds; 63 with recorded organic conversions). Key findings: organic search delivered 39 times more sessions than AI assistants (2,825,594 versus 72,095). AI referrals equalled 2.6% of organic sessions. AI sessions converted 1.18x as well as organic search (7.18% versus 6.08%), with wide variation by customer type: B2C 1.51x, Mixed 1.46x, B2B Services 1.21x, B2B Products 0.62x. AI referral volume rose 36-39% across customer segments after ChatGPT began showing brands as clickable inline links in May 2026. Growth in AI referrals correlated with growth in organic traffic (r = +0.53 after controls); no support was found for a "billboard effect" on direct traffic. A 5% improvement in organic performance would add more conversions than doubling AI referral traffic. The record contains the paper (method, results, limitations) and the aggregated dataset with site identities removed and replaced by an index number and customer type. The web version is at https://www.seoworks.co.uk/downloads/ai-referral-traffic-benchmark/
Overview
CoChat Summary
TL;DR
A 12-month GA4 benchmark across 95 websites finds AI assistant referrals are small relative to organic search (2.6% of organic sessions) but convert somewhat better overall (7.18% vs 6.08%). The conversion advantage is not universal — it reverses for B2B product sites — and organic search remains the far larger conversion opportunity.
Key findings
- Organic search delivered roughly 39x more sessions than AI assistants: 2,825,594 organic sessions versus 72,095 AI referral sessions; AI referrals equalled 2.6% of organic sessions.
- Across the 63 sites with recorded organic conversions, AI sessions converted 1.18x as well as organic (7.18% vs 6.08%), varying by customer type: B2C 1.51x, Mixed 1.46x, B2B Services 1.21x, B2B Products 0.62x.
- AI referral volume rose 36–39% across customer segments after May 2026, when ChatGPT began showing brands as clickable inline links; August 2026 reached 3.86% of organic sessions (organic advantage 26x, down from 45x in September 2025).
- Growth in AI referrals correlated positively with growth in organic traffic (r = +0.53 after controls), and the authors found no support for a "billboard effect" lifting direct traffic. A 5% improvement in organic performance would add more conversions than doubling AI referral traffic.
Methods
- Design: 12-month multi-site observational benchmark using Google Analytics 4 data, September 2025 to August 2026, from 95 properties; 66 met volume thresholds (100+ AI sessions and 1,000+ organic sessions) and 63 of those had recorded organic conversions and form the conversion analysis.
- AI referral sessions were identified by matching the referring domain against a partial regular expression on GA4 session source: `bard|chatgpt|openai|claude|copilot|gemini|perplexity|grok`. Conversions are GA4 key events configured by each business (e.g., submitted form, completed sale); conversion rate is key events divided by sessions.
- Aggregation: every percentage is derived once from summed counts within a group (sessions summed, conversions summed, then divided); per-website rates are not averaged. Median site ratio and leave-one-out ranges are reported as sensitivity checks. Sites are classified as B2C, B2B Services, B2B Products or Mixed, with identities replaced by an index number and customer type in the released dataset.
Limitations
- The paper's own limitations section is not included in the provided text, so its stated limitations cannot be reported here.
- Evident from the text: 29 of 95 properties failed the volume thresholds and 3 of the 66 threshold-meeting properties lacked recorded organic conversions, so the conversion findings rest on a smaller, threshold-filtered subset.
- Evident from the text: conversions depend on each business's own GA4 key event configuration, which is not standardised across sites; AI traffic identification relies on a source-string regex that may not capture all AI assistant referrals; and the correlation between AI and organic growth (r = +0.53) is associational, not causal.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
0 Valid · 0 Not valid · 0 Not checked · 12 Own results and interpretations
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Across the 63 sites with recorded organic conversions, AI sessions converted 1.18 times as well as organic (7.18% versus 6.08%), but the advantage was not universal”
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“AI referrals equalled 2.6% of organic sessions overall.”
- Also stated in: § Abstract · plain
“with AI referrals equal to 2.6% of organic sessions”
§ 3. Results > 3.1 Volume: organic search dominates
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § Abstract · plain
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Across the sample, organic search delivered 39 times more sessions than AI assistants (2,825,594 versus 72,095), with AI referrals equal to 2.6% of organic sessions.”
- Also stated in: § 3.1 Volume: organic search dominates · plain
“Over the 12 months, the sample received 2,825,594 organic sessions and 72,095 AI referral sessions, a ratio of roughly 39 to 1.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § 3.1 Volume: organic search dominates · plain
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“AI referral volume rose 36-39% across customer segments after ChatGPT began showing brands as clickable inline links in May 2026.”
- Also stated in: § 3.1 Volume: organic search dominates · plain
“from May 2026, when ChatGPT began making brands clickable inline, AI sessions rose 36-39% across customer segments”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § 3.1 Volume: organic search dominates · plain
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“we found no support for a 'billboard effect' in which AI mentions lift direct traffic”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
2 more claims
Create a free account to see all 12 claims and every source verdict.
- It may state two separate claims as one. 3
- Probably not a substantial claim. 2
Citations
Cites
Still being checked.
Cited by
Still being checked.
Metadata
- Authors
- Works, T. S., & Foster, B.
- Venue
- The SEO Works
- Publisher
- The SEO Works
- Published
- 2026-09-29
- OpenAlex
- W7214780540
- Cited by
- 0
- References
- 0
- FWCI
- 0.00
- Type
- journal article
- Language
- en
- Open access
- green
- doi.org
- Licence
- Locations
- Zenodo (CERN European Organization for Nuclear Research) (cc-by)
- Topics
- Social Sciences › Business, Management and Accounting › Marketing › Digital Marketing and Social Media
- Physical Sciences › Computer Science › Information Systems › Web visibility and informetrics
- Physical Sciences › Computer Science › Information Systems › Information Retrieval and Search Behavior
- Keywords
- ChatGPT, organic search, search engine optimization, Google Analytics 4
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-11-11, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-11-11 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.
