ThriveStack citedbyAI search attribution learning guideIntroductionLive demoBook a demoStart free
Lesson 7.3 · How attribution models work

AI attribution models:
which one credits AI search fairly

Last-touch models undercredit AI search, because buyers often meet an AI answer first and convert later through branded search or direct. First-touch credits AI more fairly but only for visits you can see. The practical answer is to report AI-sourced and AI-influenced revenue side by side and label the second as an estimate.

23x

higher conversion rate for AI search visitors than organic search visitors on one site

Ahrefs, single site, last 30 days to Jun 2025
4.4x

value of an AI search visitor compared with an organic visitor, based on conversion rate

Semrush, Jul 2025
83%

higher revenue per visit for AI traffic than non-AI traffic on U.S. retail sites

Adobe, Sep 2026 data, Oct 29, 2026
90%

higher conversion rate for AI visits than non-AI visits on U.S. retail sites

Adobe, Sep 2026 data
ThriveStack citedby

The AI answer gets 0% to 100% of the credit depending on the attribution model you pick

Worked example: one buyer, four touches, and how each model splits the credit.

First touch100%Last touch100%Linear25%25%25%25%Time decay6.7%13.3%26.7%53.3%Position based40%10%10%40%1 AI answer read2 Google brand search3 Webinar4 Direct demo request

Worked example. Time decay uses weights 1, 2, 4 and 8.

How to read it. Same journey, five answers. First touch gives the AI answer everything. Last touch gives it nothing. Report the model next to the number.

In ThriveStack citedby Sourced and influenced revenue are shown as two lines, so a model choice cannot hide the AI gap. See it for your site →

Each attribution model gives AI search a different amount of credit.

The model you pick changes the answer. Here is what each one does with an AI visit that came first.

Last-touch hides AI search, and first-touch shows it only when the first visit was tracked
ModelHow credit is splitEffect on AI credit
First touchAll credit to the first recorded visitCredits AI when its visit was recorded
Last touchAll credit to the final visitOften credits branded search or direct, not AI
LinearEqual credit to every visitGives AI a share if recorded
Position basedMost credit to first and lastGives AI credit when it is first
Data drivenModel estimates each touchNeeds volume and clean data

First touch plus a survey answer is the simplest fair setup.

Use first-touch for AI-sourced revenue because it is a count. Use the survey answer for AI-influenced revenue because it is the only view of buyers who never clicked.

Use first touch for
  • AI-sourced revenue
  • Because it is a count you can verify
Use the survey answer for
  • AI-influenced revenue
  • Because it is the only view of buyers who never clicked

Do not let a model hide the gap between sourced and influenced.

A single blended number looks tidy and hides the uncertain part. Two lines, one labelled estimate, is more honest and easier to defend.

One blended number
  • Looks tidy
  • Hides the uncertain part
  • Hard to defend when finance asks how it was built
Two lines
  • One counted, one labelled estimate
  • Shows the gap openly
  • Easier to defend

Studies agree that AI visitors convert well, and they disagree on how well.

Ahrefs reported that 0.5% of its visitors came from AI search and drove 12.1% of signups, a conversion rate about 23 times higher than organic search. It is one site, and Ahrefs says so. Semrush estimates 4.4 times across its study. Adobe measured an AI conversion lead of 90% on U.S. retail sites in September 2026, against a 9% gap in the other direction in August 2025.

The range is wide because sites, products and periods differ. Pick one conservative multiple for planning and let your own data replace it.

~23x

AI visitor signup rate against other visitors, on one site (0.5% of visitors, 12.1% of signups)

Ahrefs, Jun 16, 2025. One site.
4.4x

value of an AI visitor against organic

Semrush, Jul 21, 2025. Vendor study, method not explained.
+90%

AI conversion rate against other traffic in Sep 2026, after 9% worse in Aug 2025

Adobe, Oct 29, 2026

The model you choose decides who gets credit, so run two side by side.

Semrush notes that its projections include buyers who arrive with no referrer, buyers who type your URL after an AI answer and buyers who search your brand later. A last click model credits none of them to AI.

Run last click and first touch together for 90 days. The gap between them is the size of the question mark you should show your finance team.

  1. 1

    Run a last-click report

    This is the default in most analytics tools.

  2. 2

    Run a first-touch report beside it

    Same period, same data.

  3. 3

    Compare both models for 90 days

    The gap between them is the part of AI value that last click hides.

How ThriveStack citedby helps

ThriveStack citedby reports AI-sourced and AI-influenced pipeline separately, so no single model hides the gap.

Build my AI attribution report →

AI attribution models, answered.

Which attribution model is best for AI search?

Report first-touch AI-sourced revenue and survey-based AI-influenced revenue side by side.

Why does last-touch undercount AI?

Buyers often convert later through branded search or direct, and those visits get the credit.

Do I need data-driven attribution?

Not to start. It needs volume and clean data, and AI visits are still a small share.

Sources for this lesson

  1. Ahrefs: 0.5% of visitors drove 12.1% of signups (Patrick Stox, Jun 16, 2025) Single site
  2. Semrush: impact of AI search on SEO traffic (Rachel Handley, Jul 21, 2025) 500+ topics
  3. Adobe: AI traffic drives 83% higher revenue per visit (Oct 29, 2026) 1 trillion visits

Figures were checked against each page in Nov 2026. Vendor studies are marked by their sample. Charts marked sample data are illustrations.

Questions this lesson answers

See all 10 questions in this cluster
  • First-touch attribution for AI searchP0
  • Last-touch AI referral attributionP0
  • Multi-touch AI search attributionP0
  • AI-assisted conversion attribution modelsP0
  • Account-based AI search attributionP0
  • Time-decay attribution for AI discoveryP0
  • Position-based attribution for AI referralsP0
  • Data-driven attribution for AI trafficP0
  • AI search attribution model comparisonP0
  • How to avoid double-counting AI revenueP0
Go deeper
MapHover to open