Out of Ad Ideas? Write New Google and Meta Ads From What Already Works

Your best ad has been live for months, its frequency keeps climbing, and every replacement you shipped this quarter lost to it. Another brainstorm won't fix that. The same team, reading the same dashboard, starts from the assumptions that produced the ads now wearing out, so it writes them again in new outfits.

Fresh Google and Meta ads come from extracting, not inventing: the reasons your winners won, the words your customers already use, and the claims you can prove. Each platform then imposes its own discipline on how those ideas are written, and your test budget, not a quota, decides how many you launch each week.

Where do new ad ideas come from?

Most teams that feel they've tried everything have actually tried one angle in a dozen executions. An angle is the reason someone should care: price, time saved, status, relief from a specific frustration. An execution is how you say it: the hook, the format, the creator. Suppose you swap a testimonial for a UGC video and the video for a carousel. If all three make the same promise to the same person, the customer has seen one ad three times.

That's hard to spot from inside an account. Every new idea inherits the assumptions that produced the first winner. A group brainstorm rarely breaks this, because everyone in the room is reading the same data.

So change the input rather than pushing harder. Four sources reliably produce ad copy ideas that a brainstorm won't:

• Why your winners won. CPA and ROAS tell you whether an ad worked, not why. Split each winner into the parts the platform measures separately:

- scroll-stop rate for the hook - hold rate for the story - outbound click-through for the offer

A strong hook with weak click-through is telling you to keep the opening and test a new reason to buy. A loser with a strong hold rate and a dead hook has a good story that nobody reached.

• What customers already say. The best ad ideas from customer reviews usually sit in the three-star ones, because they explain what someone expected rather than simply praising. Other sources do the same job:

- Your Google Ads search terms report shows the problem as people framed it before they'd heard of you. - The sentence a prospect uses on a sales call to explain why they started looking often beats anything the team wrote. - Churn reasons, inverted, give you the objection to remove upfront.

• Other categories. Competitors converge on the same benefits in the same order because they watch each other. Borrow structure, not creative, from ads that have nothing to do with yours.
• The offer itself. A new guarantee, trial structure or bundle creates an angle that didn't exist yesterday. Warren Jolly's analysis of the 2025 Cannes Lions Grand Prix winners found that several changed the product rather than the ad. AXA, for example, rewrote its insurance policies to include domestic violence coverage.

This step should produce phrases, not concepts. "Saves time" is a concept every competitor already uses. "I stopped doing payroll on Sunday nights" is an angle only your customer could hand you.

Keep a running library of what your winners share, so the pattern outlives whoever spotted it. The full method, including the diagnostic metrics and where to read customer language, is in where good ad ideas come from when you've tried everything.

Everything below depends on this step. The platform rules shape how an idea is written, but none of them can rescue an idea that was never specific.

What should a Facebook ad's first line say?

On Meta, the first line of primary text is the only copy most people read. Feed placements truncate the rest, so the line works less as an introduction than as a caption for the creative beneath it. It should name the reader's specific situation, or the outcome they want, and address them directly. It also has to make sense with nothing beside it. It shouldn't lead with your brand, a feature or a vague category of pain.

Specificity does most of the work. "Struggling with reporting?" names a topic. "Your Monday starts with three exports and a spreadsheet that never reconciles" names a morning the reader has actually had.

The second line also filters the audience for you. Anyone who doesn't recognise that morning scrolls past, which is what you want from a placement you pay for. It also sounds like something a customer said, tightened, and that is usually where strong first lines come from.

Formats such as a direct question, "POV:" or a numbered list do work, which is exactly why the feed is full of them. The format is only a container, and the stop comes from the detail inside it.

Questions carry a particular risk, because a yes/no question invites a no. "Want better ads?" is easy to dismiss. "Still rebuilding the same client report every Friday?" earns an involuntary yes.

Treat published Facebook ad copy examples and hook rankings with care. The comparisons in circulation tend to classify a few hundred ads from the Meta Ad Library and count how long each stayed live. They treat that longevity as performance, but an ad can stay live simply because nobody paused it. Those tallies are a reason to test an opener type in your own account, not proof that it will win there.

Your own account is the better swipe file. Read the first lines of the ads that held their spend longest, side by side. Patterns in phrasing often show up more clearly than patterns in creative.

Placement changes the execution, not the principle. In Stories and Reels, the primary text gets little attention, so the hook moves into the creative as a text overlay or the first line of the voiceover. Write one hook and adapt it per placement. If the video already opens on that line, the primary text should add the outcome or the proof rather than repeat it.

The full breakdown, including when to name the product and how the register shifts for B2B buyers, is in what the first line of a Facebook ad should say.

How do you fit headlines into 30 characters?

Google search demands the opposite discipline. Responsive search ad headlines are capped at 30 characters each, with spaces and punctuation included, and you can supply up to 15. Google shows up to three of them at a time, in combinations it chooses. Those limits are long-standing, but platform specifications change, so check Google's current help page before you build a large batch.

The consequence is that each headline has to make sense alone and next to any other headline you've written. A trimmed sentence fails that test. Split "Get a Free Quote for Your Home Insurance Today" across two fields and both halves read like errors when shown out of order.

Write single ideas instead, each with one job:

• Keyword match, so the ad visibly mirrors the search.
• Offer, such as a price, discount, free trial or free delivery.
• Proof, such as reviews, years trading or accreditations.
• Differentiator, meaning what the next ad on the page doesn't do.
• Action, meaning what happens when they click.

A few headlines should carry the keyword. The rest should carry reasons to click without it, so Google has genuinely different lines to combine. Pin sparingly, because every pin removes combinations the system could otherwise test.

Compression is mostly deletion. "Get a Free Quote for Your Home Insurance Today" is 46 characters, and "Free Home Insurance Quote" is 25. The searcher loses nothing, because ads are read as offers and "get" and "your" are implied. A few more rules do most of the remaining work:

• Numerals beat words, as in "Free Shipping Over $50".
• Short words beat long ones.
• Noun phrases collapse. "24/7 Emergency Plumbers" carries what took 58 characters in sentence form.

What you shouldn't do is abbreviate into code, or let the squeeze push a claim past what's true. "Lowest Prices" fits easily and is hard to defend.

Notice what the constraint strips out. In 30 characters there's no room for scaffolding, so only what the writer knows about the business survives. That is exactly where generic AI copy fails. A model with no knowledge of your positioning fills the space with the category's stock phrases, such as "Quality Service You Can Trust", because those are the shortest plausible lines it has seen.

An agent drafting headlines needs the brand's tone of voice, positioning and ICP definitions in front of it. SproutMe Knowledge holds those per workspace, so short copy doesn't default to generic copy.

Before uploading, count in bulk with `=LEN()` in a spreadsheet. Then read random draws of three headlines aloud to catch repetition and contradiction. The full technique list, plus the traps in double-width scripts and dynamic keyword insertion, is in how to write Google Ads headlines that fit in 30 characters.

How do you word claims so ads get approved?

A distinct, specific, well-compressed line can still fail review, and the words alone are rarely the reason. Google's enforcement guidance lists what feeds its decisions: ads, accounts, websites, user complaints, regulatory rulings and third-party sources. So the unit under review is the claim, checked against the page it sends people to and the business behind both.

"Certified installer" can be well written and still be disapproved, because the certificate is nowhere on the site. That's also why two near-identical ads can get different verdicts.

Disapprovals for claims usually fall into four patterns:

• Claims about who you are, such as partnerships, endorsements or licences you can't show.
• Offers you can't deliver.
• Outcomes the evidence doesn't support, such as "guaranteed results" or "clinically proven".
• Claims about the reader, on Meta.

That last pattern is where good hook-writing advice collides with policy. The first-line method says to name the reader's situation and address them directly. Meta's personal attributes rule says copy that asserts or implies something about the viewer can fail, even under broad targeting. That covers their health, finances, age or beliefs. "Struggling with debt?" is a perfectly shaped hook and a likely rejection.

The two reconcile once you separate a situation from an attribute. A Monday morning spent reconciling spreadsheets is a situation the reader can recognise without being told anything about who they are. Debt, a diagnosis or an age bracket is an attribute.

In sensitive categories, describe the product and let the reader self-select. "Debt advice that starts with a free review" does the job "Struggling with debt?" can't do safely. Check Meta's own policy text for the current list of covered attributes.

The stakes differ by policy. That matters for anyone who treats review as something to iterate against.

• Misrepresentation. At the time of writing, Google gives at least seven days' warning before suspension, with strikes up to three per policy.
• Unacceptable business practices. Google's Unacceptable business practices policy is classed as egregious. Accounts are suspended on detection, without prior warning.

Claims about credentials, partnerships or health get verified before they ship. They never get tested in live traffic to see what passes.

The practical rules are short:

• Write every claim so the landing page proves it within one click.
• Prefer the checkable claim to the superlative. "Same-day dispatch on orders before 2pm" beats "fastest delivery anywhere".
• Keep qualifiers next to the claim.

Then keep an approved-claims list, recording the evidence, an owner and a last-checked date for each claim. That list closes the loop with the first section. Proven claims are part of what already works, and new copy built from them starts pre-cleared.

The four patterns, the appeal process and what to do with a rejected ad are covered in why ads get rejected and how to word claims safely.

How many new ads should you test each week?

Volume is where most teams overcorrect once the ideas start flowing. The right weekly number is however many tests your budget can bring to a verdict within that week. For most accounts, that means one or two new concepts, each run as three to five clearly different executions in its own ad set. Your cost per result sets the number, not a quota.

The arithmetic is simple and unforgiving. Every new Meta ad set starts in a learning phase, and it only stabilises after collecting enough optimisation events within a week. To size your tests:

1. Multiply your target cost per result by the number of results a test cell needs before you'd trust its verdict.
2. Divide your ring-fenced testing budget by that figure.

Take a hypothetical where one cell needs $2,000 and you have $3,000 for testing. You can afford one new concept. Launch four and each gets $750, so you finish the week with four inconclusive results instead of one useful one.

If the maths doesn't fit, you have two honest options. Either optimise test cells for a cheaper event higher in the funnel and confirm any winners later, or test less often. Whatever you choose, leave each cell running for a full seven days so weekday and weekend behaviour both show up.

The unit you count matters as much as the number:

• A concept is a different argument for buying: a new angle, problem or buyer.
• A variation is the same argument told differently, such as a new hook or thumbnail.

Concepts find winners, and variations extend them. A team shipping 20 near-identical ads a week has run one test, and probably not a readable one. That's why the angle-versus-execution distinction from the first section isn't academic. It decides whether your testing programme learns anything at all.

Cadence also differs by platform. Paid social has fast feedback and fast fatigue, so it suits a weekly rhythm. Google search is slower. A responsive search ad tests headlines inside a single ad, and a low-volume ad group can take far longer than a week to say anything about a new asset. A monthly review of asset performance fits there better.

Ads stuck in review also lose their slot in the week. That's one more reason to settle claims before the batch is written.

Finally, judge the programme by hit rate, not launch count. Of the ads you launched, what share went on to win and take spend?

• If hit rate is healthy and you still need winners, raise volume.
• If hit rate is falling, more launches will only fail faster. The problem is the ideas going in, which sends you back to your winners and your customers' words.

The budget maths, kill rules and weekly routine are laid out in how many new ad variations to test each week.

Conclusion

Running out of ad ideas almost always means running out of inputs. Your next ad copy that converts comes from the ads converting now. Start with the metrics that explain why a winner won, the exact words customers use before they know you, and the claims you've already proven.

Each platform then shapes those ideas differently. On Meta, that means a recognisable situation in the first line. On Google, it means one idea with one job in each 30-character headline. Policy decides which of those ideas can be said at all, and on Meta, situations pass where personal attributes don't.

Test budget decides how many ideas you can judge in a week. Hit rate tells you whether to write more or write differently.

See how SproutMe Companion answers questions like which hooks and concepts took share of spend last quarter, grounded in your live account data.