Products SEO & GEO Agent Paid Ads Agent Website Agent Conversion Agent
How it works Pricing FAQ Blog Book a call

AI fake data in marketing: spot it and protect your decisions

Updated · 7 min read

Short answer Fake data reaches a small business three ways: AI tools inventing facts, bots filling your forms, and bot traffic or clicks in your reports. Check every number an AI gives you against its original source, block form bots, and judge ads by booked jobs, not form fills. Otherwise your ad system's AI learns from the fakes.

Want help applying this to your business? Talk it through on a free 30-minute call.

Book a call

Where does fake data get into your marketing?

Through three doors. Each needs a different check.

DoorWhat it looks likeQuick check
AI tools invent factsA statistic, quote, study or competitor figure in a draft that nobody can traceOpen the original source and find the exact number
Bots fill your formsLeads with fake phone numbers, random names, no reply to any callCall five recent leads; count the real ones
Bots visit or clickTraffic or clicks jump with no rise in calls or bookingsCompare traffic against booked jobs, week by week

Fake reviews are a fourth door, with their own rules: see AI-generated reviews.

The damage is rarely the fake item itself. It's the decision built on it: an ad budget moved, a page rewritten, a claim published under your name.

How do I spot made-up facts from AI tools?

Treat every number, quote and source an AI tool gives you as unchecked until you've seen it on the original page. AI chat tools guess when they don't know. A 2025 research paper on why they make things up puts it plainly:

"...large language models sometimes guess when uncertain, producing plausible yet incorrect statements." (Kalai et al.: Why Language Models Hallucinate)

A "hallucination" is that guess: a confident answer that's false. The paper's point is that the tools are trained in a way that rewards guessing over saying "I don't know". So the error looks just as sure as a true answer.

Check it in three steps:

  1. Ask for the source. If the tool can't name the page, cut the figure.
  2. Open the page yourself. Find the exact sentence. Made-up citations exist; so do real pages that say something different.
  3. Check the date and place. A 2019 US figure is not a 2026 Canadian one.

Red flags: round, dramatic numbers ("73% of customers..."), a study with no author or year, a quote from a named person you can't find anywhere else.

Why it matters more in ads

Once a claim is in your ad or on your site, it's your claim, whoever wrote it. In the US, the FTC's guide for small businesses says:

"The law requires that advertisers have proof before the ad runs." (FTC: Advertising FAQ's: A Guide for Small Business)

In Canada, the Competition Act bars performance claims "not based on an adequate and proper test", and says "the proof of which lies on the person making the representation" (Competition Act, section 74.01). An AI-invented "cuts energy bills 40%" fails both.

How do I spot fake leads?

Look at the leads, not the count. Bot and junk leads tend to share signs:

  • Phone numbers that don't connect, or the same number on several leads
  • Names or messages that are random letters, or copy the form labels
  • Many submissions within minutes, often overnight
  • Messages selling you something (SEO, loans) instead of asking for your service
  • Free email addresses that bounce

Then close the door:

FixWhat it does
Hidden "honeypot" fieldA form field people can't see; bots fill it, so those entries are dropped
Cloudflare TurnstileChecks for bots "without showing visitors a CAPTCHA" (Cloudflare Docs); the free plan has unlimited checks and works without moving your site to Cloudflare
Google reCAPTCHA v3Gives each visit a score: "1.0 is very likely a good interaction, 0.0 is very likely a bot" (Google for Developers)
Google Ads lead formsGoogle suggests "OTP Verification", a one-time code sent to the person, to screen leads (Google Ads Help: About lead form assets)
Meta lead adsDefault forms are "optimized for the volume of leads"; the Higher Intent setting adds a step where a person "reviews and confirms their answers" (Meta for Developers: Lead Forms for Ads)

Each extra step costs some real leads too. Add one fix, then compare real leads before and after.

Why fake leads make your ad AI buy more fake leads

Google's and Meta's automatic bidding learn from what you tell them is a success. Google says:

"Google Ads Smart Bidding refers to bidding strategies that use Google AI to optimize for conversions or conversion value." (Google Ads Help: About Smart Bidding)

A "conversion" is whatever action you told the platform to count. If every form fill counts and half are bots, the system learns to find more people like the bots. Your cost per lead can drop while booked jobs drop too.

The fix is to count what's real:

  • Count a booked job or a qualified call, not a raw form fill, where you can.
  • Send real results back to Google. Offline conversion imports let you measure "what happens in the offline world after your ad results in a click or call to your business" (Google Ads Help: About offline conversion imports).
  • Review lead quality before you let any AI tool move budget: see AI ad spend optimization.

Bot traffic and invalid clicks in your reports

Some filtering is done for you, but not all, and you can't see how much.

  • Google Analytics 4: "traffic from known bots and spiders is automatically excluded." But "you cannot disable known bot traffic exclusion or see how much known bot traffic was excluded" (Google Analytics Help). Unknown bots still get through.
  • Google Ads: Google calls invalid clicks those that "don't represent a genuine interest in your business", and says "You won't be charged for invalid clicks or impressions." If found after billing, you get a credit "where appropriate and possible" (Google Ads Help: About invalid traffic). The Invalid traffic card on your Overview page shows the summary.

Signs that bots are in your numbers anyway: a spike from one city or country you don't serve, visits that last zero seconds, or a jump in sessions with no change in calls. Judge your site by key events like calls and bookings: see GA4 key events for a small business.

What to do when fake data already got in

  1. Find the start date. When did leads, traffic or clicks change without a matching change in jobs?
  2. Stop the spread. Pause the campaign or form bringing junk. Stop any AI tool from acting on that data.
  3. Fix the success count. Make sure ad platforms stop counting the fake conversions as wins.
  4. Clean the records. Mark or remove fake leads in your customer list so reports and follow-ups skip them.
  5. Close the door. Add a form check, a source check for AI drafts, or a person's approval before changes go live. For who approves what, see human oversight in AI marketing.

Example

Illustrative cases with invented numbers.

A plumber's form fills double (US dollars)

Google Ads shows 80 leads this month, up from 40, at $25 each instead of $50. Calls back reach 12 real people, about the same as before. Bots found the form. The plumber adds Turnstile, switches the counted conversion to booked jobs, and spends the same $2,000 a month on fewer, real leads.

A dental clinic's AI blog draft

An AI draft says "64% of adults skip cleanings because of cost", credited to a dental association. The office manager searches the association's site and can't find the figure. It's cut, and replaced by the clinic's own payment plan details.

A Vancouver renovator's traffic spike (Canadian dollars)

Visits triple in a week, mostly from a country the business doesn't serve, with near-zero time on site. Bookings are flat. The owner doesn't raise the C$1,500 monthly ad budget on the strength of the "growth", and judges the month by booked estimates instead.

FAQ

Can AI tools detect fake data for me?

Partly. Turnstile and reCAPTCHA screen forms, and Google filters known bots. Only a person opening the source can confirm an AI-written statistic.

Is it illegal to publish an AI-made statistic?

A false or unproven claim is the problem, whoever wrote it. The FTC wants proof before the ad runs; Canada wants performance claims tested first.

Do I get money back for bot clicks on my ads?

On Google Ads, clicks flagged as invalid aren't charged, and ones caught after billing are credited where possible. Unflagged clicks are billed.

How often should I check lead quality?

Weekly is enough for most small businesses: call a handful of new leads and compare leads to booked jobs. Check sooner if lead volume jumps.

Different for you?

Keep learning

NextClaude and Gemini News: What Small Businesses Should LearnWhat the Claude Cowork file deletions and Gemini's Personal Intelligence mean for a small business, and the habits that keep your files and client data safe.

Ask us one question

One line is enough. We reply by email.

Sources

  1. Kalai, Nachum, Vempala and Zhang: Why Language Models Hallucinate (arXiv, September 2025)
  2. FTC: Advertising FAQ's: A Guide for Small Business
  3. Justice Laws Website: Competition Act, section 74.01
  4. Google Ads Help: About Smart Bidding
  5. Google Ads Help: About offline conversion imports
  6. Google Ads Help: About lead form assets
  7. Meta for Developers: Lead Forms for Ads
  8. Cloudflare Docs: Turnstile
  9. Cloudflare Docs: Turnstile plans
  10. Google for Developers: reCAPTCHA v3
  11. Google Analytics Help: Known bot-traffic exclusion
  12. Google Ads Help: About invalid traffic