Writing Ad Copy With AI: Prompt Patterns That Work

AI genuinely speeds up producing headlines, descriptions, and CTAs for Google Ads and Meta Ads — but only with the right prompt. Type a flat “write me ad copy” into ChatGPT or Claude and you’ll generally get low-click-through, generic sentences anyone could write: “quality service,” “best price.” The difference lives in the prompt itself. This article is part of the AI Automation Guide series; I walk through the 4 prompt patterns I use on my own accounts, how I fit them to Google Ads character limits, and the recurring mistakes in AI-written copy — with real examples.

Prompt Patterns That Work

All four patterns rest on the same logic: instead of a blank page, you give AI concrete constraints and real business data. You can reuse the same pattern over and over just by swapping the business information.

a) Benefit-focused headline pattern

Prompt: “You’re a Google Ads copywriter. Write 8 headlines for the following business, each no more than 30 characters (including spaces). Business: a same-day teeth-whitening clinic in Istanbul. Target audience: time-constrained people aged 30-45 due to work. Each headline should contain a concrete benefit (duration, result, convenience); don’t use unprovable adjectives like ‘best,’ ‘quality,’ ‘leading.’ Output as a numbered list, with the character count in parentheses at the end of each line.”

Example output:

• Same-Day Teeth Whitening (24)
• White Teeth in 60 Minutes (26)
• Done on Your Lunch Break (25)
• Book Today, Done Today (23)

b) Objection-handling pattern

Prompt: “For the same business, list the 3 hesitations a customer thinks through before deciding on this service (e.g. uncertain pricing, whether the result lasts, whether it hurts). Then write a Google Ads description, no more than 90 characters, that addresses all of these hesitations at once. Don’t use phrases like ‘guarantee,’ ‘100%,’ ‘definite result.’”

Example output: “Price Is Clear Before You Book, No Hidden Fees. Appointment Confirmed Instantly. (79)”

c) Urgency/scarcity pattern

Prompt: “For this business, write a description based on a real constraint (not fake urgency) — for example, a limited number of weekly appointment slots, or a campaign ending on a specific date. No more than 90 characters. Don’t use phrases that could trip Google Ads policy, like ‘100%,’ ‘guarantee,’ ‘don’t miss out.’”

Example output: “Only 12 Appointments Left This Month. Calendar Filling Fast, Book Now. (72)”

d) Competitive differentiation pattern

Prompt: “Without using a competitor’s name (Google Ads doesn’t allow it), write an indirect differentiation sentence: start with a phrase like ‘unlike other clinics’ or ‘different from most places,’ then add a concrete difference (appointment speed, pricing transparency, equipment). No more than 30 characters, headline format.”

Example output: “No Waiting, Direct Booking (27)”

Fitting Google Ads Character Limits

On Google Ads, headlines max out at 30 characters and descriptions at 90 (Responsive Search Ads allow up to 15 headlines and 4 descriptions). AI models are bad at counting characters — they work token-based, not character-based. Skip the limit and you’ll get 45-50 character headlines; Google either rejects them or cuts them off mid-word, and a truncated headline both lowers click-through and looks broken.

What I do is simple: at the end of every prompt I add a clear formatting instruction — “no more than 30 characters, including spaces, write the character count next to each line.” Then I verify the output not by eye but with a real character counter (like Google’s own RSA preview), one line at a time. Don’t trust AI’s own character count — it getting the count wrong is a common occurrence.

I also have AI prepare performance reports on the same accounts — I walk through that process in Preparing Your Google Ads Report With AI.

Common Mistakes in AI Copy

Publishing AI’s first draft as-is is the most common mistake I see. Three recurring problems:

• Generic phrasing: “the best,” “quality service,” “industry leader,” “unmatched experience” — these don’t differentiate any business from another, and AI defaults to them.
• Drifting from brand voice: AI’s default tone tends to be either too formal or too “marketer-y”; it may not match the business’s actual voice — casual, technical, playful, whatever it is.
• Google Ads policy risk: phrases like “guarantee,” “100%,” “definite result,” “scientifically proven,” or “free” when it isn’t actually free carry the risk of rejection or, on repetition, account-level restriction — especially in healthcare, finance, and appointment-based services.

That’s why my rule is firm: I never publish AI’s output directly. Every line is either edited by hand or cut. Even a good prompt doesn’t produce publish-ready copy on the first pass — it produces the edited version on the second.

An Example From My Operation: What Changed When the Prompt Changed

At one of the accounts I manage (a mid-sized business in the services sector — I can’t share the name), we’d gone through a period of publishing AI’s first output almost as-is: 5 headlines full of “quality service,” “expert team,” “best price.” Three weeks later, click-through rate was noticeably below the account’s own average.

We rewrote the same headlines using the objection-handling pattern — versions with concrete duration and numbers — and ran them in an A/B test against the old ones. The edited version raised click-through rate by 38%. The difference wasn’t using AI vs. not using AI — it was whether we published the first output or the second, constrained-prompt output.

I use the same logic on the content side — I walk through having AI prepare a social media posting calendar in AI Social Media Calendar.

You Can Ask Me to Do This

You can ask me to do this: reviewing your existing ad copy and rewriting it with the patterns above, then A/B testing it — work I generally run entirely remotely, hourly, taking 2 hours to 2 weeks. Scope is on the Ad Creative Package page.

Frequently Asked Questions

Can I publish AI-written ad copy as-is?

I wouldn’t recommend it. AI’s first output usually contains generic phrasing and incorrect character counts; it needs at least one round of manual editing and verification with a character counter.

Which AI tool gives better results for ad copy?

The prompt structure matters far more than the tool. ChatGPT, Claude, or Gemini can all produce similarly good results using the same patterns; the real difference is in the constraints and examples you give.

Do these prompts work for Meta Ads too?

The core logic is the same (benefit, objection-handling, urgency, comparison) but character limits differ — Meta has no fixed limit on primary text, though shorter is still recommended. Just adjust the character limit in the prompt to match the platform.

Can AI-generated bold claims put my Google Ads account at risk?

Yes, especially unprovable phrases like “guarantee,” “100%,” “definite result” can lead to rejection and, on repetition, account-level restriction. These need to be filtered out before publishing.