Self-Reported Attribution: Adding 'ChatGPT' to Your Forms

Self-reported attribution means asking people how they found you, usually with a "How did you hear about us?" field, and adding options such as ChatGPT, Perplexity, Gemini and "Another AI assistant". It captures influence that referral analytics misses: the buyer who asked ChatGPT for tool suggestions, then later searched your name. Set up properly, it takes an afternoon and produces one of the most persuasive GEO metrics for leadership.

Key takeaways

• Analytics sees the last click; buyers remember where they discovered you.
• Use a dropdown with AI options plus optional free text.
• Store the answer in your CRM against the contact and deal.
• Report it alongside, not instead of, analytics data.

Why does analytics miss AI influence?

Even a well-built GA4 AI channel only captures the first row. Self-reported data covers the rest, with its own biases.

Which form should get the field?

Prioritise forms closest to revenue:

1. Demo or sales contact requests
2. Free-trial or signup forms
3. Quote requests
4. Newsletter signups (lower priority, though often higher volume)

Ecommerce teams often use a post-purchase survey on the order confirmation page instead.

How should you design the field?

Label: "How did you first hear about us?" ("First" pushes people towards discovery rather than the last step.)

Dropdown options (illustrative):

• Google or another search engine
• ChatGPT
• Perplexity
• Gemini
• Microsoft Copilot
• Claude
• Another AI assistant
• Social media
• Podcast or YouTube
• Colleague or friend
• Review site
• Event
• Other

Optional follow-up: a free-text box such as "Anything more? (e.g. what you searched or asked)".

Randomise the order of options if your form tool allows it, to reduce position bias, or keep a consistent, logical grouping and accept the bias. Don't put AI options first just to inflate them.

How do you route answers into your CRM?

1. Create a contact property, for example self_reported_source, with the same option values.
2. Map the form field to it.
3. Optionally copy it to the deal or opportunity at creation, so it stays attached to revenue.
4. Create a separate text property for free-text answers.
5. Check a few test submissions end to end.

Keep values stable. Renaming "Other AI" to "Another AI assistant" mid-year splits your data unless you migrate old values.

How do you clean free-text answers?

Code free-text answers monthly into the same categories. Typical patterns: "chatgpt", "chat gpt", "GPT", "asked AI", "AI search". Keep a mapping table so coding stays consistent between people. Flag ambiguous answers ("Google AI") as their own category rather than guessing whether the person means AI Overviews or Gemini.

How do you analyse and report it?

Illustrative, not real data: in a quarter, 6% of demo requests select an AI assistant while GA4 attributes 1% of demo-request sessions to AI referrals. The gap is the interesting finding: AI influence exceeds what analytics shows.

Should sales ask the question too?

Forms capture one moment; conversations capture more. Add a short prompt to discovery-call notes: "How did you come across us, and what else did you look at?" Sales teams often hear things like "ChatGPT suggested you and two others", which is rich detail no form will capture. Log the answer in the same CRM property, or a separate "sales-reported source" property, so you can compare the two.

Ask sales to note the competitors the AI suggested alongside you. Over a quarter, those notes show which brands you're grouped with, which is useful context for your recommendation rate tracking.

Worked example

Illustrative. A 40-person SaaS company adds an optional "How did you first hear about us?" dropdown to its demo form, with five AI options and a free-text box. It maps the field to a HubSpot contact property and copies it to deals at creation. After one quarter, the team codes 60 free-text answers, finds that most AI-related answers mention comparing tools, and briefs the content team to strengthen comparison pages. The quarterly report shows self-reported AI share, GA4 AI referral key events and AI-sourced pipeline side by side, each with its sample size.

What biases should you watch for?

• Recall bias: people misremember, especially for long journeys.
• Option bias: listing AI assistants makes them more likely to be chosen.
• Social desirability: some people pick "Colleague" by default.
• Small numbers: a monthly sample of 40 leads swings widely; report quarterly or with ranges.

None of these make the data useless. Track it consistently and look at trends.

Common mistakes

• Making the field mandatory on a conversion-critical form without testing.
• Using one generic "AI" option that hides which assistant.
• Not copying the source to deals, so you can't connect to revenue.
• Changing option labels without migrating values.
• Treating it as precise attribution rather than directional evidence.

Checklist

• [ ] Field added to the highest-value forms
• [ ] AI assistant options listed individually
• [ ] Optional free-text follow-up
• [ ] CRM property mapped and copied to deals
• [ ] Monthly coding of free text
• [ ] Quarterly report alongside GA4 AI sessions

Where to go next

• GA4 custom channel group for AI referrals
• AI visibility KPIs for marketing teams

Sources

• Google Analytics Help, Custom channel groups
• Google Analytics Help, About key events