Testing demand in 2026: what to get right before you run an intent test

In 2020 I wrote a guide to intent testing: put a real proposition in front of a real audience with an ad, send the people who click to a simple page, and see how many take the next step. It's still the cheapest honest signal you can get before you build anything.

I've been on both sides of this. Faraday, the intent testing tool I wrote that guide for, went on to become Tapkit, which was acquired by noissue in 2022.

The method hasn't changed. The rulebook has. Between new UK consumer law, new cookie rules, US privacy signals and the EU AI Act, a demand test that was fine in 2020 can now cost you a fine, an ad account or your reputation. None of it makes testing harder if you set it up properly. Here's what I'd check before launching one in 2026.

1. Be honest that it doesn't exist yet

The classic "fake door" test shows a product that isn't built and measures who tries to walk through. The test isn't the problem. Pretending is.

In the UK, advertising must not mislead about availability, and the CAP Code names availability explicitly. Since April 2025 the CMA can fine businesses directly for misleading consumers, up to 10% of global turnover, and its first fines under the new regime came in 2026.

In the US, if you take money for something you can't ship yet, the FTC's Mail Order Rule applies: you need a reasonable basis for the shipping date you state, or for shipping within 30 days if you don't state one.

Say "coming soon" or "join the waitlist" on the page, not "buy now".
Don't take payment for a product you can't deliver. If you take deposits to test willingness to pay, say exactly what they're for, when you'll ship and how refunds work.
When someone clicks through, tell them straight away that it isn't available yet and what happens next.

2. Tracking needs consent, even in a test

Intent tests live on measurement: an ad pixel, a conversion event, an analytics tag. That's exactly where the rules tightened.

In the UK, the Data (Use and Access) Act changed PECR on 5 February 2026. Cookies used solely for statistics about your own site, and cookies that remember how the site looks, no longer need consent, as long as you offer a simple, free way to opt out. Advertising pixels don't qualify: a cookie that does analytics and ad targeting still needs consent. And the maximum fine for getting cookies or marketing consent wrong went from £500,000 to £17.5m or 4% of global turnover.

If you advertise with Google in the UK or EEA, you also need Consent Mode v2 sending the user's choice, or you lose remarketing and conversion measurement for anyone who hasn't consented.

In the US, a growing list of states require you to honour Global Privacy Control, the browser signal that says "don't sell or share my data". Regulators check whether tracking actually stops when the signal is on, not just whether there's a banner.

A consent banner that actually blocks ad tags until people say yes.
First-party, statistics-only analytics for the test itself. In the UK that no longer needs consent, and it's enough to read a funnel.
Honour GPC if you're running US traffic.

3. Social proof has to be real

Landing pages love "2,000 people on the waitlist" and a row of five-star quotes. Both are now explicitly regulated.

In the UK, fake reviews and reviews that hide that they were paid for or incentivised have been banned outright since 6 April 2025, under Schedule 20 of the DMCC Act. In the US, the FTC's Consumer Review Rule bans fake reviews and testimonials, with civil penalties of over $50,000 per violation.

If a number is on the page, it's true and you can show where it came from. No invented quotes, and no AI-generated "customers".

4. Label AI-generated people

It's never been easier to make ten versions of an ad with an AI presenter. Since 2 August 2026, the EU AI Act's transparency rules require that AI-generated content showing realistic people is disclosed as AI-generated. A brand that chooses a synthetic presenter for its own ad is responsible for the label.

If a person in your test ad doesn't exist or didn't say it, label it, especially when the ad reaches EU audiences.

5. Price tests show the real price

Testing willingness to pay is one of the most useful things an intent test can do. But the CMA's early fines under its new powers were about pricing: drip pricing and automatic opt-ins for paid extras. If you show a price, show the total, including mandatory fees.

6. Waitlist data is real personal data

The email address someone gives your fake door is personal data. Collect only what you need for the test, say what you'll do with it, and if you want to market to that list later, get clear consent at sign-up. If you don't launch, tell people and delete the list.

7. The platforms have rules too

Meta's ad standards require the landing page to work and to match what the ad promotes. A test page that leads nowhere, or switches to a different offer, gets ads rejected and can put the account at risk. Build the test page as a real page.

None of this makes intent testing worse. If anything, the honest version tells you more: people who join a waitlist knowing the product isn't ready are a stronger signal than people tricked into clicking "buy".

This post is about practice, not legal advice. If you're running tests at scale or in regulated sectors like finance or health, talk to a lawyer.