How to Use ChatGPT for Google Ads: Keywords, Ad Copy and Negatives

Most people use ChatGPT for Google Ads in the laziest possible way. They type "write Google ads for my clinic", copy whatever comes out, paste it into the ad and wonder why half the headlines get cut off or why the ad says things the business never promised.

ChatGPT is genuinely useful for Google Ads, but only for certain jobs. It's great at sorting a messy keyword list, giving you fifteen headline angles in a minute and spotting patterns in a search terms report. It's bad at counting characters, it doesn't know your prices or your area, and it has no idea what Google's ad policies will reject this week.

So in this guide I'll show you the workflow I'd use for a small business in Pakistan or India: what to hand to ChatGPT, the exact prompts, and the checks you must do yourself before anything goes live. The limits and policy points come from Google Ads Help pages, which I checked on 31 October 2026. If you haven't set up a campaign yet, start with my Google Ads for beginners guide and come back here when you're writing keywords and ads.

What ChatGPT is good at (and what it isn't)

Here's how I split the work:

Keep that table in mind and you'll avoid most of the mistakes I see.

One more thing before the prompts: don't paste customer names, phone numbers or anything private into ChatGPT. You don't need it. Search terms, keywords and your own business facts are enough.

Step 1: Give ChatGPT a fact sheet first

Every bad AI ad I've seen has the same cause: the AI was guessing. So before asking for anything, write a short fact sheet and paste it at the start of the chat. Something like this:

Business: [name], [service], in [city and areas served]. What we sell: [3-5 services or products, with real price ranges if you show prices]. Proof we can honestly claim: [years in business, certifications, warranty, Google rating only if true]. Offer right now: [offer, or "no offer"]. How people contact us: [call, WhatsApp, booking form]. Things we must never say: [for example "cheapest", "guaranteed results", competitor names]. Landing page: [the exact page the ad will send people to].

That last line about "never say" saves a lot of editing. If you don't offer free home visits, tell ChatGPT that up front, or it will happily invent them.

It also helps to know what your competitors already say. Google's free Ads Transparency Center shows the ads other advertisers have run in your country, so you can spot the offers everyone repeats and the angles nobody uses. My Google Ads Transparency Center guide shows how to search it. Just don't paste competitors' ads into ChatGPT and ask it to rewrite them. Use them to decide what makes you different.

Step 2: Turn a messy keyword list into tight ad groups

Collect your raw keywords first. Use the phrases customers actually type and say on WhatsApp, Google autocomplete and Keyword Planner. My keyword research guide using free tools covers that part.

Then paste the list into ChatGPT with this prompt:

Here is a list of keywords for a [business type] in [city]. Group them into ad groups where every keyword in a group means the same service. Give each group a short name. Put keywords that show job seekers, students, DIY or free intent in a separate list called "possible negatives". Don't add new keywords and don't remove any. Output as a table: ad group, keyword.

Two instructions in that prompt matter. "Don't add new keywords" stops it from inventing phrases nobody searches. And the "possible negatives" list gives you a head start on Step 4.

Then review the groups yourself. ChatGPT sometimes lumps "AC installation" together with "AC repair" because they look similar. They're different jobs with different landing pages, so they deserve separate ad groups. A good rule: if two keywords need a different landing page, they belong in different ad groups.

For match types, I still start beginners on phrase and exact match, as I explain in my Google Ads beginners guide. ChatGPT can format your list for you: ask it to wrap each keyword in quotes for phrase match or square brackets for exact match.

Step 3: Write responsive search ads that actually fit

Google's responsive search ads page sets the rules you have to write within:

• 3 to 15 headlines, up to 30 characters each
• 2 to 4 descriptions, up to 90 characters each
• 2 path fields, up to 15 characters each

Google mixes and matches these in different orders, and its help page says headlines and descriptions may appear in any order, so every line must make sense on its own. Google also suggests writing the first three headlines as if they'll appear together.

Here's the prompt I use:

Using the fact sheet above, write 15 headlines for a Google responsive search ad for the ad group "[ad group name]". Rules: maximum 30 characters each including spaces. Each headline must make sense on its own and next to any other. Include the keyword "[main keyword]" in at least 3 headlines. Write 3 about the service, 3 about the offer, 3 about trust, 3 calls to action and 3 about the area. Only use facts from the fact sheet. After each headline, write its character count in brackets.

Then ask for four descriptions with a 90-character limit in the same way.

Never trust ChatGPT's character count

This is the part most people skip. ChatGPT will often write "(28)" next to a headline that's actually 32 characters long. Google Ads will show you a counter when you paste, but it's faster to check everything at once in Google Sheets:

1. Paste the headlines into column A.
2. In B1, type =LEN(A1) and drag it down.
3. Anything over 30 gets shortened by you, not by asking ChatGPT again and again.

Do the same for descriptions with a limit of 90. Google's page also notes that every character in double-width languages like Chinese, Japanese and Korean counts as two, so if you ever write ads in those languages, the limits are effectively halved.

Pinning: use it sparingly

If something must appear in every ad, like a licence number or a legal disclaimer, Google says you need to pin it to headline position 1, headline position 2 or description position 1. But Google also says pinning reduces the number of combinations it can test and isn't recommended for most advertisers. My approach: pin only what's legally or commercially essential, and let the rest rotate.

Example: AC repair in Karachi

Here's the kind of set you're aiming for (invented business, check your own facts):

• AC Repair in Karachi (20)
• Same-Day AC Service (19)
• AC Gas Refill and Repair (24)
• Book on WhatsApp Today (22)
• Technicians in Clifton & DHA (28)
• Fixed Price Before We Start (27)

Description: "Tell us the AC model on WhatsApp and we'll share the visit charge before we come." (81)

The numbers in brackets are real character counts. Notice there's no "Best in Karachi" or "100% Guaranteed". You can't prove those, and Google's Misrepresentation policy doesn't allow ads that mislead people or omit important information about products, services or pricing.

Don't let AI write in "ad-speak"

ChatGPT loves shouting. Ask for "catchy" headlines and you'll get "UNBEATABLE DEALS!!" and "B.E.S.T Prices". Google's Editorial policy covers capitalisation that isn't used correctly, gimmicky repetition and unusual spacing, and its Punctuation and symbols policy gives examples like "flowers!!", "fl@wers" and "F.L.O.W.E.R.S." as not allowed. The Editorial policy also lists entering a phone number in the ad text as a separate issue, so use a call asset instead.

I add one line to every prompt now: "Use normal sentence or title case, no repeated punctuation, no symbols as letters, no phone numbers."

Step 4: Use ChatGPT to build negative keywords from real searches

This is my favourite use, because it saves real money. Once your campaign has run for a week or two, go to Insights and reports > Search terms. Google's search terms report page explains that this shows the actual searches that triggered your ads. You can download the data from the report.

Then paste the search terms (not your whole account) into ChatGPT:

These are search terms that triggered my Google ads for [service] in [city]. We do NOT offer: [list]. Suggest negative keywords that would block irrelevant searches without blocking good ones. For each, say why, and suggest whether it should be broad, phrase or exact match negative. Group them into themes like jobs, DIY, free, other cities, other services.

Now the human part. Before you add anything, remember how negatives work. Google's negative keywords page says:

• Negative keywords don't match close variants. Blocking "job" doesn't block "jobs". You have to add singular, plural and synonyms yourself. Ask ChatGPT to list both forms.
• Casing and misspellings are handled for you, so you don't need "JOBS" or common typos.
• Negative broad match (the default) blocks a search only if it contains all your negative words, in any order.
• Negative phrase match blocks searches containing those words in that order.
• Negative exact match blocks only that exact search, with no extra words.

The classic AI mistake is over-blocking. ChatGPT might suggest "price" as a negative because some searches were "AC repair price list pdf". But "AC repair price" from someone in Karachi could be your best lead. Block "pdf", not "price". Read every suggestion and ask: would a real customer ever search with this word?

If you run several campaigns, put the shared junk words (jobs, salary, course, free download) into a negative keyword list so you don't repeat the work. My beginners guide covers list limits.

Step 5: Check trademarks before you use any brand name

AI doesn't think about trademarks, so you have to. Two common situations for small businesses:

• You're a reseller or repair shop using a brand name, like "Samsung AC repair" or "Daikin service".
• You want to target a competitor's name.

Google's Trademarks policy says Google won't restrict using trademarks as keywords, but it may restrict trademarks in the ad text when a trademark owner complains, for example in ads from a direct competitor or ads that use the trademark in a confusing or misleading way. For resellers and repair services, Google looks at whether the landing page is primarily about selling or facilitating the sale of that brand's products or services, and whether the ad and page make it clear that you're a reseller or an informational site.

In plain words:

• Bidding on a competitor's name as a keyword isn't restricted by Google, but putting their name in your headline can be.
• If you repair Daikin ACs, say "We repair Daikin ACs", not "Daikin Official Service", unless you really are authorised.
• Never let ChatGPT write "official", "authorised" or "certified" unless it's true and you can show it.

Google also says it gives a warning at least 7 days before suspending an account for a Trademarks policy violation. Don't let it get that far.

Step 6: Know what Google's own AI already does

Before you spend hours generating copy, know that Google Ads now has its own text tools. Google's text customization page explains that this setting, available through AI Max for Search campaigns, generates extra headlines and descriptions from your landing page, existing ads and keywords. The same page says that from September 2026, campaigns using text customization are being upgraded to AI Max automatically.

That's another reason your landing page matters. Google's page tells you to make sure your website content is accurate and not misleading, because the generated text draws on it. If your page says "free home visit" and you've stopped offering it, Google's AI could put that in your ads. Keep the page accurate, and check the ad-level asset report to remove anything that isn't true. The page also mentions text guidelines, such as words to exclude or themes to avoid, so if you turn this on, add your "never say" list there too.

My view: use ChatGPT to write a strong, honest set of your own headlines, then decide whether to test Google's automation once you have conversion data.

A 20-minute weekly routine with ChatGPT

1. Download last week's search terms and ask ChatGPT for negative ideas. Add the good ones after checking.
2. Look for new keyword themes in the same report. Good searches you're not targeting might deserve their own ad group.
3. Check asset performance in Google Ads and replace the weakest one or two headlines with new ChatGPT drafts.
4. Run every new line through the LEN check and your "never say" list.
5. Click your own landing page on a phone. If the ad promises something the page doesn't show, fix the page or the ad.

If you also make image or video ads, my guide to using AI for ad creatives follows the same rule: AI for variety, humans for facts. And if you're still choosing an assistant, I compared ChatGPT, Gemini and Claude for marketing work.

You'll find more guides like this on my blog, the Google Ads calculators on my Google Ads and SEO page, and more about me on my home page.

FAQ

Can ChatGPT write Google Ads for me?

It can draft keywords, headlines and descriptions quickly, but you should give it your real business facts, check every character count yourself and review the copy against Google's policies before publishing.

What are the character limits for Google responsive search ads?

Google allows up to 15 headlines of 30 characters each, up to 4 descriptions of 90 characters each and two path fields of 15 characters each. You need at least 3 headlines and 2 descriptions.

Why does ChatGPT get Google Ads character counts wrong?

AI chat tools often miscount characters, especially with spaces and punctuation. Paste your lines into Google Sheets and use the LEN formula, or rely on the counter in Google Ads, before you publish.

Can ChatGPT find negative keywords?

Yes. Download your search terms report and ask ChatGPT to group irrelevant searches into themes. Then check each suggestion, because it can block useful words like "price", and add singular and plural forms yourself.

Can I use a competitor's brand name in Google Ads?

Google's Trademarks policy says it doesn't restrict trademarks used as keywords, but it may restrict a trademark in your ad text after a complaint, especially from a direct competitor. Keep competitor names out of your headlines.

Does Google Ads have its own AI for ad copy?

Yes. Text customization, available through AI Max for Search campaigns, generates extra headlines and descriptions from your landing page and existing ads. Keep your landing page accurate and review what it creates.

Keep reading

• Google Ads for beginners in Pakistan and India: your first Search campaign on a small budget
• Keyword research for beginners using free tools
• How to use AI to create ad creatives that convert
• ChatGPT vs Gemini vs Claude for marketing work: which to use when
• 25 ChatGPT prompts for social media managers
• Google Ads Transparency Center: how to see competitors' Google ads for free
• See all posts on the blog
• Back to my home page