01 the business
Group of private dental clinics
6 clinics, 110 staff
Ads, social posts and web pages are written by staff, freelancers and AI tools. The final check depends on whoever is free. Words like 'painless' and 'best results' slip through without evidence, which is a real risk for a healthcare provider.
02 before
The problem
Copy gets written by different people, freelancers and AI tools, and the final check depends on whoever is free. Words like guaranteed, cure or best slip into ads without proof. In regulated sectors that can mean a complaint to the advertising watchdog, and in any business it erodes trust.
- No single place where copy is reviewed.
- Health and price claims go live without the evidence on file.
- No record of who approved what, if a complaint arrives.
03 the build
What the AI does
- Step 1: Gives your team one place to paste or upload copy, or picks up drafts automatically from Google Docs, WordPress or HubSpot.
- Step 2: Checks tone, banned words, product names and formatting against your brand guide.
- Step 3: Highlights claims about results, prices, health, savings or comparisons and asks for the evidence behind each one.
- Step 4: Checks the copy against the advertising rules you load for your markets, such as FTC, ASA or ACCC guidance.
- Step 5: Returns a marked-up version with suggested safer wording and keeps a record of every check and decision.
paused: waiting for a person to approve
How it was put together
- Created one place to paste or upload copy, and picked up drafts from the website and email tools.
- Checked tone, banned words and product names against the brand guide.
- Highlighted claims about results, prices, health or comparisons and asked for the evidence behind each.
- Kept a log of every check, the suggested wording and the final decision.
04 human in the loop
Where a person stays in charge
The tool advises and never publishes. A named reviewer, and your compliance or legal adviser for regulated claims, signs off flagged items before the copy goes live.
- The tool advises and never publishes.
- Health claims always go to a clinician and the compliance lead for sign-off.
- The rule sets for each market are loaded and maintained by the compliance lead.
05 rollout
From kickoff to live in 4–5 weeks
- Weeks 1–2Turn the brand guide and advertising rules into a checklist with the compliance lead.
- Weeks 3–4Check the last three months of published copy to calibrate.
- Week 5Make the check a required step before anything is published.
06 after
What changes
Risky claims are caught before customers or regulators see them. Reviews take less back and forth because writers see the problems and suggested fixes in one place, and you keep a clear record of what was checked.
Assume 60 pieces of copy a month, three rounds of back-and-forth review at 15 minutes each, falling to one round.
60 × 3 × 15 minutes = 45 hours a month of review today. 60 × 1 × 15 minutes = 15 hours with marked-up drafts.
What to measure
- Share of published copy that went through the check.
- Risky claims caught per month.
- Review rounds per piece of copy.
- Complaints about advertising.
07 evidence
What the research says
The FTC's 2024 rule bans fake reviews, including AI-generated ones, undisclosed reviews by company insiders and threats used to suppress negative reviews.
US Federal Trade Commission, Final rule banning fake reviews and testimonials (opens in a new tab) · 2024
NIST's framework organises AI risk work into four functions, Govern, Map, Measure and Manage, with a profile for generative AI added in July 2024.
NIST, AI Risk Management Framework 1.0 and Generative AI Profile (opens in a new tab) · 2023
08 lessons
Mistakes to avoid
- Treating the tool as the approver. It flags; people decide.
- Loading rules once and forgetting them. Guidance changes.
- Checking only ads. Social posts and web pages carry the same risks.
09 questions
Common questions
- Which advertising rules can it check against?
- Whatever your compliance lead loads, such as FTC guidance in the US, the CAP code in the UK or ACCC guidance in Australia.
- Does it check reviews and testimonials too?
- It can flag testimonials used in ads. Fake or undisclosed insider reviews are banned under the FTC's 2024 rule.