Why Most AI-Generated Ad Copy Is Mediocre
I have been managing Google Ads campaigns for over 15 years, across legal, medical, automotive, and retail clients. When AI writing tools started getting good enough to draft ad copy, I jumped in immediately. And I quickly discovered the most common mistake people make: they ask the AI to write their ads and then use whatever comes out.
The result is predictable. Generic headlines. Weak CTAs. Copy that sounds like every other ad in the category. If you are a personal injury attorney, you do not want ads that sound like every other personal injury attorney. You want copy that makes someone stop scrolling and click.
Here is the framework I have developed after two years of using AI for Google Ads copy across dozens of campaigns.
Step 1: Build Your Brief Before You Open Claude or ChatGPT
The quality of AI ad copy is almost entirely determined by the quality of your brief. Before you open any AI tool, write down:
- The specific service or product being advertised
- The target audience and their primary pain points
- The competitive differentiators — what makes you different from the other ads on that page
- Any proof points: years in business, number of clients, specific credentials, awards, guarantees
- The desired action and any offer or incentive
- Any compliance constraints (especially important for legal, medical, and financial clients)
This brief takes 15 minutes to write and dramatically improves everything that follows.
Step 2: The Master Prompt for Google Ads Headlines
Here is the prompt structure I use with both Claude and ChatGPT:
You are an expert Google Ads copywriter. I am advertising [service] to [target audience] in [location]. Our key differentiators are: [list]. Our offer is: [offer]. Character limits: headlines are 30 characters max, descriptions are 90 characters max. Write 25 headlines and 10 descriptions that emphasize different angles including: urgency, social proof, specific benefits, problem agitation, and direct offers. Mark each angle used.
The key elements: you are specifying the output format (quantity and character limits), telling it the angles you want covered, and giving it enough context to differentiate. Generic prompt equals generic output. Specific prompt equals usable output.
Step 3: The Angle Expansion Technique
After getting your initial batch, use follow-up prompts to go deeper on the angles that resonate:
- Write 10 more headlines specifically emphasizing [specific differentiator]
- Rewrite these 5 headlines to be more urgent without being salesy
- Write 5 headlines that address the objection a prospect might have about [specific concern]
This iterative approach gets you to a library of 60 to 80 headlines per campaign, which gives you real testing data over time.
Claude vs ChatGPT for Google Ads Copy
I have tested both extensively for ad copy specifically. My findings:
- Claude tends to write more nuanced, less generic copy on the first pass. For sensitive verticals like legal and medical, Claude is more careful about compliance language.
- ChatGPT with GPT-4 is faster and the first drafts are often punchy and direct. For straightforward retail and service categories, it is slightly faster to get usable copy.
- For volume testing — generating 80 plus variants — both are roughly equivalent in speed.
My recommendation: use Claude for your primary copy development and ChatGPT for iteration and variation. They produce different angles on the same brief, which expands your testing pool.
Compliance Review: Non-Negotiable
For legal, medical, financial, and healthcare clients, all AI-generated ad copy must go through human compliance review before launch. AI does not know your state bar rules, your FDA compliance requirements, or your specific licensing constraints. I have a compliance checklist for each regulated vertical that every ad goes through before it goes live.
The Testing Framework
With AI, you can now generate enough variants to run a statistically valid headline rotation test at launch rather than adding variants manually over months. Our standard process for new campaigns: 20 headlines, 6 descriptions, pinned CTAs where appropriate. We let Google’s Responsive Search Ad optimization run for 30 days, then pull the asset report and cut underperformers. The AI generates the next round based on what is working.
Real Results
Across campaigns where we have implemented this systematic AI-assisted copy process compared to our previous process, we see average CTR improvements of 15 to 30 percent after three months of testing. That is a meaningful conversion volume increase without any increase in budget.
If you want help implementing this process for your Google Ads campaigns, see our PPC and Google Ads services or contact us for a campaign review.
