Your analytics can count clicks from ChatGPT, but it cannot count the person who read a Gemini answer on their phone, remembered your name, and called you two days later. For many businesses, that second person is the more common AI-influenced customer. The simplest way to find them is also the oldest: ask.
Self-reported attribution means asking leads and customers how they heard about you, in a form field or on a call, and recording the answer. Adding AI assistants to that question is one of the cheapest, fastest ways to see AI search's real influence. This guide covers how to add AI options to forms and sales calls, how to word the question so answers are useful, and how to combine the results with analytics data.
It complements software-based tracking such as marketing attribution models, which only see what is clicked.
Key Takeaways
- Self-reported attribution captures AI influence that analytics misses, such as copied links, app hand-offs and delayed visits.
- Use an open "How did you first hear about us?" question, optionally paired with a short list that includes "AI assistant".
- Name examples (ChatGPT, Gemini, Claude, Copilot, Perplexity) so people recognise the option.
- Record answers in your CRM next to analytics source data and compare the two every month.
- Treat the results as directional: memory is imperfect, and people often credit the last thing they remember.
Why analytics alone undercounts AI
When an AI assistant passes a referrer, analytics can recognise it. Google Analytics added a dedicated channel for that in May 2026.
“Google Analytics now provides a dedicated way to measure and analyze traffic originating from popular AI assistants.”
— Google Analytics Help, What’s new in Google Analytics (13 May 2026)
But the channel depends on a recognised referrer. Visits from apps that drop it, links people copy, and later visits through branded search or typed addresses all show up as Direct or Organic Search. We explain those gaps in why AI traffic shows up as Direct. A form question catches the customer regardless of the route they took.
Designing the question
Wording matters more than most teams expect. Compare the common formats.
| Format | Pros | Cons | Best for |
|---|---|---|---|
| Open text only ("How did you first hear about us?") | Unprompted, honest answers; reveals specific tools and prompts | Needs manual or AI-assisted coding | B2B, high-value leads |
| Dropdown only | Clean data, easy to report | Order and wording bias answers; misses detail | High-volume ecommerce |
| Dropdown plus "tell us more" text box | Structured data with detail | Slightly longer form | Most businesses |
| Multi-select ("Where have you seen us?") | Shows multiple touchpoints | Harder to report; less clear on first touch | Research surveys |
| Post-purchase survey | No friction on the conversion form | Lower response rate | Ecommerce checkouts |
Wording tips
- Ask about the first touchpoint or the most helpful one, and be consistent, since they measure different things.
- Label the option clearly: "AI assistant (e.g. ChatGPT, Gemini, Claude, Copilot, Perplexity)". Many people do not call these "AI search".
- Keep "Google search" separate, and consider "Google AI answer at the top of search" as its own option if your audience would recognise it.
- Randomise the order of dropdown options where your form tool allows it, keeping "Other" last.
- Make the field optional on short lead forms if requiring it hurts conversions; test it.
Step-by-step: adding AI to forms and sales calls
- Audit current fields. Find every place you ask the question: lead forms, booking tools, checkout, onboarding and sales scripts.
- Standardise options. Use the same list everywhere so data can be combined. Include AI assistant, Google search, social media, referral from a person, review site, event, podcast, email and other.
- Add a free-text follow-up. If someone chooses AI assistant, ask which one and what they asked, if they remember.
- Capture analytics context automatically. Use hidden form fields to store the session source, medium, campaign and landing page. Our guide to UTM parameters explains the values involved. OpenAI says ChatGPT adds
utm_source=chatgpt.comto referral links, so a hidden field that stores the landing URL can catch it. - Brief the sales team. Add one question to discovery calls: "Before you contacted us, where did you research options?" Train reps to log the exact tool named.
- Store it in the CRM. Create a dedicated property for self-reported source and another for the detail. See our guide to using a CRM for marketing.
- Code the open answers. Each month, group free-text replies into categories. AI can help draft the coding, but review it by hand.
- Report monthly. Track the share of new leads and customers who name an AI assistant, and the revenue attached.
Combining self-reported and analytics data
The real value comes from comparing what people say with what analytics recorded. Each combination tells you something.
| Self-reported answer | Analytics source | Likely story |
|---|---|---|
| AI assistant | AI Assistant channel or chatgpt.com | Clean, confirmed AI click |
| AI assistant | Direct | App hand-off, copied link or typed address after an AI answer |
| AI assistant | Organic Search (branded query) | Read an AI answer, then searched your name |
| Google search | Organic Search | Could include AI Overviews or AI Mode; analytics cannot split them |
| Friend or colleague | Direct | Word of mouth, possibly a shared link |
| AI assistant | Paid search | AI raised awareness, an ad captured the click |
Note the "Google search" row. Google's documentation says AI Overviews and AI Mode traffic is reported inside Search Console's Web search type, so customers who say "Google" may also have seen an AI answer.
The "AI assistant plus Direct" and "AI assistant plus branded search" rows are the hidden AI influence that click-based tools miss. They also explain why we treat branded search as a signal of AI visibility. When you map them across the customer journey, you often find AI acting early, as research, rather than as the last click.
Reporting it
- Show "AI-assisted leads" as the union of analytics AI sessions that converted and self-reported AI leads, with the overlap noted.
- Track the ratio of self-reported AI leads to analytics AI leads over time. A rising ratio suggests more hidden influence.
- Feed the numbers into your AI search KPIs and your AI search business case.
Common mistakes
- Hiding AI under "Other". If the option is not visible, few people will type it in.
- Using jargon. "Generative engine" means nothing to most customers. Name the tools.
- Changing options often. Every change breaks your trend line. Plan the list once and review it yearly.
- Treating answers as precise. People misremember. Use the data for direction and proportion, not exact credit.
- Collecting more than you need. Keep the question short and covered by your privacy notice, and do not ask for sensitive details.
- Never comparing with analytics. The two sources are most useful together, as part of a wider AI search ROI calculation.
Related Guides
- Branded Search as a Signal of AI Visibility
- Which Pages Get AI Traffic? Landing Page Analysis for AI Referrals
- Is AI Referral Traffic Higher Quality? How to Test It on Your Own Site
Frequently Asked Questions
Should I add ChatGPT to my "How did you hear about us?" form?
Yes. Add an option such as "AI assistant (e.g. ChatGPT, Gemini, Claude, Copilot, Perplexity)" and a follow-up text field for detail.
Is self-reported attribution reliable?
It is directional rather than exact. People forget or credit the most recent touchpoint, but it captures AI influence that analytics cannot see.
Should the field be required?
Test it. Required fields give more data but can reduce form completions. Many businesses make it required on high-value forms and optional elsewhere.
How do I combine survey answers with GA4?
Store source, medium and landing page in hidden form fields, then compare them with the self-reported answer in your CRM.
Which AI tools should I list?
List the assistants your customers are likely to use, such as ChatGPT, Gemini, Claude, Copilot and Perplexity, and review the list once a year.
Conclusion: ask, record, compare
Adding AI assistants to "How did you hear about us?" takes an afternoon and quickly shows how much AI shapes your pipeline. Pair it with analytics, report it monthly and you will have a far more honest picture than clicks alone. Need help improving forms and tracking? Our conversion rate optimization service can help, or request a free quote.
