“How did you hear about us?”: how to measure the customers AI sends you

Where to ask “how did you hear about us?” (forms, checkout, CRM and support), which answer options to use, how to include ChatGPT and other AIs, how to match the answer with GA4 and HubSpot, and how far self-reported attribution can be trusted.
Written with AI, reviewed by humans

In this article

See how your brand shows up in AI answers

Book a demo and find out what to fix first, with a GEO specialist.

Book a demo

Reading time: 11 minutes

To measure the customers who come from AI, ask “how did you hear about us?” on forms, at checkout and in the CRM. Then add how did you hear about us options that name ChatGPT and other AIs, so the survey captures them. After all, the chat happens off your site and rarely leaves a trace in GA4.

In short

  • The “how did you hear about us?” question records the AI chat that didn’t lead to a click, which GA4 counts as direct or organic search.
  • Ask it where the lead already gives you data: contact form, checkout, first sales call and support.
  • In the how did you hear about us options, include “AI recommendation”, a “Which one?” field and an open field, and sort the answers every month.
  • Read the result as a range, next to the GA4 AI channel, because the answer depends on the customer’s memory.

Why ask “how did you hear about us” if GA4 already tracks the source?

Because GA4 only records what happens on your site. Someone who asks ChatGPT for suggestions, reads the brand name and then searches Google shows up as organic search. Someone who types the address becomes Direct. So asking how did you hear about us is the most direct way for a company to record the chat without a click.

The guide to AI traffic in GA4 shows the AI Assistant channel and what it leaves out. In addition, the pillar on SEO and GEO ROI explains why no attribution model gives credit to direct visits. The only exception is a path made up entirely of direct visits. The details of each model are in the guide to the attribution model for AI.

The CRM has the same blind spot. According to HubSpot help, a contact’s traffic source can be AI Referrals, with chatgpt.com in the drill-down. However, in our reading of the help example, this value also depends on a click on the assistant’s link.

In US retail, the problem is already visible. In the PYMNTS Intelligence U.S. Playbook, only 24% of 584 US merchants could tell when a sale came from an AI platform. It was fielded in March 2026. Visa Acceptance Solutions commissioned the report, and Visa is a payments company, so it has an interest in the topic.

How did you hear about us: illustration of a form with answer options, one marked with an AI icon, linked to a bar chart

Where should you ask how did you hear about us?

Ask how the customer heard about the company at the first point where they already share data. That means the contact or demo form, the checkout and the first support contact. This way, the question doesn’t add a new step. Then send the answer to the same CRM field, wherever it comes from.

TouchpointWhen to use itWatch out for
Contact formServices and B2BShort field, near the end
Checkout or thank-you pageEcommerceOptional, after payment
First sales callConsultative salesRep logs it in the same field
Phone, chat and supportLocal businessQuestion in the opening script

In consultative sales, the best answer often comes up in conversation. For example, the rep asks how the customer heard about the company on the first call. Then the rep writes down the exact answer in the CRM. The rep shouldn’t correct the customer or suggest options, because that changes the answer.

Should the how did you hear about us field be required?

On a contact form, a required field makes sure every lead says how they found the company. However, it also leads to rushed answers, like “internet” or “I don’t remember”. At checkout, prefer an optional field, because the purchase comes before the survey. In any case, track how many leads left the how did you hear about us survey field blank, since that rate is part of the reading.

Ask only for what you will use in the how did you hear about us survey. In the EU and UK, the GDPR asks for a specific purpose that is known to the person (Article 5(1)(b)). It also asks for data that is adequate, relevant and limited to what is necessary (Article 5(1)(c)). In California, the CCPA requires collection to be “reasonably necessary and proportionate” to its purposes, as the California Privacy Protection Agency explains in its 2024 enforcement advisory, citing Civil Code § 1798.100(c). In our reading, this means the how did you hear about us survey should say what the question is for. This is general guidance, not legal advice, so ask a privacy professional how these laws apply to your business.

Which how did you hear about us options should you list?

In the how did you hear about us survey, list six to eight options with your real channels, plus “Other: which one?”. Among the how did you hear about us options, separate search, AI recommendation, referral, social media, ads and events. Finally, leave an open field, because it reveals channels the list didn’t foresee.

Here is a sample of how did you hear about us options for the survey (illustrative, so adjust them to your channels):

  1. Google or Bing search
  2. AI recommendation (ChatGPT, Gemini, Perplexity, Copilot)
  3. Referral from a customer, friend or colleague
  4. Social media (Instagram, LinkedIn, TikTok)
  5. Ad
  6. Event, podcast or press
  7. Other: which one?

If your form allows it, shuffle the order of the options and keep only “Other” fixed at the end. That way, people who pick the first option in a rush don’t always favor the same channel. Also avoid generic options like “Internet”, which lump search, AI and social media into one answer.

Open field or closed list?

Use both. The closed list of how did you hear about us options gives numbers that are easy to add up, while the open field shows how the customer describes the path. The 2024 report from HockeyStack covers this. HockeyStack sells B2B attribution software. The data came from 8,528 responses from demo bookers at dozens of B2B companies, mostly in software, in North America, Europe and Australia. Among the forms analyzed, only two had a list. All the others used an open field, and the company recommends that format.

The open field takes work, though. In the same analysis, roughly 20% of the answers were invalid or too generic, like “online”. So set a fixed time each month to sort the open answers.

How do you add ChatGPT, Gemini and other AIs to the options?

Create a dedicated option, like “AI recommendation (ChatGPT, Gemini, Perplexity)”, and then show a “Which AI?” field. This way, the answer doesn’t get lost in “Other” or in “Google search”. Among the how did you hear about us options, this is the one that changes how you read the funnel.

Fairing sells attribution surveys, with a focus on ecommerce. It analyzed over 80,000 free-form answers to how did you hear about us surveys, and it found “ChatGPT”, “chat gpt”, “chatgbt” and simply “AI”. In addition, many people described the AI as something that recommends. Therefore, the company suggests the option “AI Recommendation (e.g. ChatGPT)”.

When you sort the answers, merge the variants into one category. For example, “chat gpt”, “GPT” and “I asked AI” all become “AI recommendation”. Next, store the assistant’s name in a second field, because ChatGPT, Gemini and Perplexity need different GEO work.

Be careful with the word “Google”. Google Search now shows AI answers, so a customer who read an AI Overview may pick “Google search”. If the open field says “Google’s AI”, classify it as AI and note the engine.

How do you match the answer with GA4 and the CRM?

Store the answer as a CRM field and send only the category to GA4, in a lead event parameter. Then compare the declared source with the tracked source, lead by lead in the CRM and by channel in GA4. As a result, each how did you hear about us survey answer becomes a number your attribution report can use.

In GA4: a lead event with the declared source

The GA4 events reference recommends the generate_lead event for forms. It already has the lead_source parameter, described as “The source of the lead.” So send the chosen category in it, such as “ai_recommendation”, and never the free text.

Next, create an event-scoped custom dimension with that parameter. According to GA4 help, the data shows up 24 to 48 hours later. The same page advises against high-cardinality dimensions, which push data into the “(other)” row. Also, free text may include an email or phone number. Google Analytics policies forbid sending personally identifiable information.

Finally, in an exploration, cross the declared source with the session channel group. That way, you see how many leads that named AI arrived as Direct or Organic Search.

In HubSpot and other CRMs

In HubSpot, create a custom contact property for the declared source. HubSpot’s help lists dropdown select among the field types, which fits the closed list. Then compare it with the original traffic source, which shows AI Referrals when there was a click in the assistant.

In other CRMs, the idea is the same: a dedicated field with the same how did you hear about us options as the survey, filled in by the form or by the rep.

Lead’s answerTracked sourceReading
AI recommendationAI Assistant or AI ReferralsAI with a click
AI recommendationDirect or organic searchAI without a click
Google searchAI AssistantCheck the classification
ReferralAnyWord of mouth

How do you read the how did you hear about us survey results?

Add up the answers once a month and calculate two rates: one over those who answered and one over all leads. The first is the ceiling, and the second is the floor. So the AI share becomes a range, not an exact number, and the report shows both sides.

Here is an example with illustrative numbers. In one month, the company got 200 leads, and 150 answered the how did you hear about us survey question. Of those, 18 named an AI. So the range goes from 9% (18 of 200) to 12% (18 of 150). At the same time, GA4 showed only 5 leads in the AI Assistant channel. The gap is the chat without a click.

Also track sales, not just leads. To do that, filter won deals in the CRM by declared source. In addition, compare the trend with the brand’s presence in AI answers, using the protocol to measure brand mentions and citations in AI. If both rise together, there is a signal, not proof.

When you present it, show the GA4 AI channel as the floor and the how did you hear about us survey range separately. The guide to digital marketing KPIs shows how these numbers become targets. In the SEO report for clients, this range sits next to the GA4 AI channel and is never added to it.

What if the number is small?

With few leads per month, combine three months before you compare. After all, one answer more or less moves the rate a lot. In small companies, read every answer that names AI, because the open field tells you what the customer asked the assistant. If organic falls while AI answers rise, the question helps you look into the organic traffic drop with the client. Still, it doesn’t prove the cause on its own.

Is self-reported attribution reliable?

It is reliable as a signal, not as an exact measure. The answer about how the customer found the company depends on their memory, on the list options and on who decides to answer. For this reason, it complements GA4 but doesn’t replace it.

The first limit is memory. In the HockeyStack report, based on B2B leads, the overlap with search was only 53% (2,646 of 4,939). That is, when the last tracked touch was a search, only 53% of answers also named search. For social media, the overlap was 60%. In other words, customers remember a different path than the one tracking records, and neither tells the whole story.

The second limit is bias:

  • List order: people in a hurry may pick the first option.
  • Who answers: people who leave the field blank may have a different profile.
  • Channel names: the customer may call an AI answer “Google”.
  • Last contact: people tend to remember the most recent channel, not the first one.

Even so, the question is worth the effort, because it is cheap and records what no tracking sees. Therefore, keep the same how did you hear about us options for months. Change one thing at a time, and always report the range next to the GA4 AI channel.

Want to know how your brand shows up in AI answers today?

Book a demo and see how your brand shows up in AI answers, with support from a GEO specialist.

Main sources

About this articleWritten with AI and reviewed by humans, with verified sources. Found something out of date? Write to the NotHuman team.

Get new NotHuman articles by email

By subscribing, you agree to receive blog articles by email and to the Privacy Policy. Unsubscribe any time.

Keep reading

Demo

Want to know how your brand shows up in AI answers today?

Book a demo and see how your brand shows up in AI answers, with support from a GEO specialist.

Book a demo

Your privacy

We use necessary cookies to make the site work and Google Analytics and Microsoft Clarity statistics cookies to improve our content. You can decline the statistics ones; in the EU, UK and Switzerland they are only turned on after you accept. Cookie Policy.

Cookie preferences

Cookie Policy