By Mehdi HassanWhy does B2B attribution break when buyers use AI search?
B2B attribution breaks in AI search because the research that builds a buyer's shortlist happens inside ChatGPT, Perplexity or an AI Overview, and most of it never produces a trackable click. When the buyer finally visits, they type your URL or search your brand name, and analytics credits Direct or branded organic for a decision made somewhere else.
Forrester reported that 94% of B2B buyers used generative AI during their last purchase. The behavior isn't a fringe case any more. Yet most attribution setups were designed for a world where a buyer clicked an ad or a search result, read a few pages and filled a form. In that world, the click trail roughly matched the decision trail. Now it doesn't.
This guide is part of our hub on marketing website analytics. It argues for a simple, two-part setup that most B2B teams can run without new software, and shows the form copy that gets honest answers.
What does the AI-era dark funnel look like?
The dark funnel is the part of a buyer's journey your tools can't see: conversations, communities, podcasts and now AI answers. In the AI version, a buyer asks a chatbot for options, reads a comparison it writes, asks follow-up questions about pricing and security, and arrives at your site already leaning toward you or away from you.
What the buyer did | What analytics records | What really drove it |
|---|---|---|
Asked ChatGPT for vendors, tapped your link in the app | Direct | AI recommendation |
Read an AI Overview, then searched your brand | Organic Search, branded query | AI Overview mention |
Asked Perplexity, clicked a cited source | Referral from perplexity.ai | AI citation (tracked) |
Heard you on a podcast, asked ChatGPT to compare you | Direct or branded search | Podcast plus AI comparison |
Clicked a retargeting ad after all of the above | Paid Social | Everything before the ad |
Only one row in that table gets tracked to the right source. A last-touch model would give the final row's deal to the retargeting ad. That's how teams end up shifting budget toward the channel that harvests demand and away from the work that created it. Our guide to tracking AI traffic in GA4 recovers the tracked row; the rest needs a different instrument.
Why use self-reported attribution alongside analytics?
Self-reported attribution captures what analytics can't: the buyer's own account of what made them reach out. Analytics captures what self-reporting can't: precise, complete data on every tracked visit. Each is wrong in predictable ways, so running both gives you a far better picture than either alone.
Pros
- Sees dark channels: AI tools, podcasts, peers, communities
- Costs one form field, not a software contract
- Reflects what the buyer thinks mattered, which is what you want to repeat
- Works even when cookies, referrers and consent banners strip tracking
Cons
- Buyers forget and simplify, so it's directional, not precise
- Some leave it blank or write something unhelpful
- Answers need light cleaning before you can report on them
- It records one memory, not the whole journey
The way to read them together: analytics tells you which tracked channels and pages convert, and the self-reported field tells you which untracked influences are bringing people to those pages. When they disagree, the disagreement is the insight. For the full setup, see our step-by-step guide to adding a self-reported attribution field.
What form copy gets honest attribution answers?
Form copy gets honest answers when the question is specific, the field is open text or has an "Other" option, and there's no pre-selected default. Asking "How did you hear about us?" is fine; asking "What made you reach out today?" often gets a richer answer, because it invites the trigger, not just the first touch.
Wording | What you tend to get | Use it when |
|---|---|---|
How did you hear about us? | First memorable source: "Google", "a friend", "ChatGPT" | You want a simple, comparable answer |
What made you reach out today? | The trigger: "saw you in a comparison", "pricing page" | You want to know what tipped the decision |
Where did you first hear about us? (e.g. ChatGPT, podcast, colleague) | More AI and dark-channel answers, because the examples prime them | You suspect AI research is undercounted |
Keep the field on demo and contact forms, where it costs the least relative to the value of the lead. On high-volume newsletter signups, it adds friction for little insight. Make it optional if your form is already long, and required if it's short and the leads are valuable.
How do you report both systems without confusing leadership?
Report both systems in one table, one column each, with a sentence explaining that they measure different things. Leadership doesn't need a reconciliation formula. They need to see where the two agree, where they don't, and what you're doing about it.
- Group self-reported answers into a short list of buckets: AI tools, Google search, referral or colleague, social, podcast or event, other.
- Pull qualified leads by original source from your CRM for the same period.
- Put the two side by side and highlight the biggest gaps.
- Write two or three sentences on what the gaps suggest, such as "Self-reported AI mentions are three times the tracked AI channel."
- Tie each gap to an action: pages to improve, sources to earn mentions on, or channels to test.
Hypothetical example: what a gap tells you
For example, suppose a B2B startup gets 40 qualified demo requests in a quarter. Its CRM, fed by hidden form fields, attributes 4 to the AI channel, 14 to Direct, 12 to organic search and 10 to paid. The self-reported field tells a different story: 13 buyers mention ChatGPT or Perplexity, many of them recorded as Direct or branded search.
The tracked AI number says AI is minor. The self-reported number says it's the largest single influence. The sensible conclusion isn't that one number is right, but that AI research is driving visits analytics can't trace, so the pages AI tools draw on (homepage, pricing, comparisons) deserve investment. These figures are illustrative, not from a real company.
What should you change once you see the gap?
Once you see the gap between tracked and self-reported AI influence, invest in the things AI tools draw on: your money pages and what other sites say about you. Don't try to force the dark traffic into analytics; use it to direct effort.
- Money pages first. In TripleDart's study of 28 B2B SaaS firms, homepages took 81.7% of AI-attributed leads, while general blog posts drew about 12,000 sessions and no leads. Make pricing, security, comparison and use-case answers easy to find and quote.
- Earn mentions. Ahrefs' study of 75,000 brands found web mentions correlated 0.664 with AI Overview visibility, against 0.218 for backlinks.
- Answer the follow-ups. Buyers ask AI about pricing, security and integrations. A clear AI trust page and visible FAQs give engines something accurate to cite.
- Connect it to revenue. Pass the self-reported answer into the CRM so you can follow it to closed deals, as in our guide to connecting website analytics to your CRM.
If you're weighing paid attribution software instead, read multi-touch attribution for B2B first. And for what happens when AI visitors do click through, see does AI search traffic convert.
