A weekly ad refresh for a fixed product list: what to automate and what not to

If you sell the same products all year, the ads for them still wear out. People see the same picture again and again, and clicks drop. The common advice is to "refresh every week". That is too blunt. A small budget may not need it, and a big one may need it sooner.

This page gives a better rule: refresh on signals, change one thing at a time, and automate the making but not the approving. It shows where Higgsfield can help and where a plain template or a spreadsheet is enough.

This page may contain affiliate links from which Blur Studio earns commissions.

The problem: ads go stale and nobody has time to remake them

The pattern is familiar. An ad does well for two weeks. Then the cost per click creeps up, the owner notices a month late, and a new version is made in a rush. A routine fixes this. You do not need a new ad for every product every week. You need a short check, a short brief and a place to record what you tried.

How to tell an ad is stale: signals, not the calendar

Creative fatigue means people have seen an ad so often that they stop reacting to it. Look for these signs:

• Frequency rising. Frequency is the average number of times one person has seen your ad. A cold audience (people who do not know you) seeing the ad 2.5 or more times in a week is a common warning point.
• Click-through rate falling. CTR is the share of people who saw the ad and clicked. A drop of 20% to 30% over several days, while frequency climbs, points to fatigue.
• Cost per result rising with no change in offer or audience.

Those 2.5 and 20 to 30% figures come from an ad-analytics vendor's blog (Superads), and the vendor does not trace them to a source. Treat them as a starting point, not a law. Your own log will tell you more than any blog after three weeks of rows.

Calendar advice varies. Segwise, another ad-analytics vendor, says a practical starting point is every 2 to 4 weeks on Meta and every few days to a week on TikTok, and it calls those guesses until your own data confirms them. Neither platform publishes a rule that I found. Use your numbers.

What to change first, and what not to touch

Change one thing at a time so you know what worked. The order that most guides suggest:

1. The hook (the first line or first second, the headline).
2. The visual (the main image).
3. The angle (the main promise, such as price, speed or trust).

For a fixed product list, the cheapest weekly change is the headline and the first image. Keep the product photo, price format and logo. That is also the easiest part to automate. A still ad that has gone stale can also become a short clip: see old ad images to new reels.

What not to change on autopilot: the offer, any claim, the legal line, and anything with a real person's face.

Do not refresh every ad in the same week. That is my own rule of thumb, not a vendor finding. Keep the current best performer running and put the new version beside it, so you can see whether the change helped before you switch anything off.

Options for a fixed product list

Cheapest first.

1. One ad by hand in a free design tool, plus a saved template. Copy the template each week and swap the headline and photo. Free, and fine for under 8 products.
2. Canva. A drag-and-drop design tool. It has a free plan and its pricing page lists limited AI ad features there. Check the vendor page for what each paid plan costs in your currency.
3. Google Ads asset tools. If you run Google ads, Google has its own creative suggestion features. Check the vendor page for what is included and for the price.
4. Creatify. An AI tool that makes video and image ads from a product link. Its pricing page (read 3 November 2026) shows Standard at $10 a month and Pro at $38 a month on annual billing, or $15 and $55 billed monthly, with a free plan of 10 credits and a watermark. Its Pro and Max plans list ad automation features, but check what each one means before you pay, because the names are broad.
5. Higgsfield Ads Studio. Makes still image ad variations from your website (for a no-designer version, see ad images with your logo and colours). Covered next.

If automation and direct publishing matter more than creative quality, Creatify's higher plans may suit you better. I would read its feature list carefully first.

Where Higgsfield Ads Studio fits

Per Higgsfield's own Ads Studio page, it builds a brand kit from your website (logo, colours, brand voice and selling points), imports your products with names, images and prices, and makes variations for each product, with headlines and calls to action. A call to action (CTA) is the button text, such as "Shop now". The page says the shapes are 1:1, 4:5 and 9:16, though Higgsfield's in-product quote tool for Ads Studio, read through its Claude connection on 2 November 2026, listed 1:1, 9:16, 16:9 and 3:4 and no 4:5, so check the ratio menu.

What that means for a weekly routine:

• It makes variants on demand. Its pages do not say it runs itself on a schedule. Higgsfield has a separate agent tool called Supercomputer with scheduled jobs on paid plans (Plus allows up to 2), but whether it can drive Ads Studio is not confirmed. Do not plan around automation Higgsfield has not documented. See the Supercomputer and Canvas guide.
• Cost is not published. No credit cost appears on the page. The cost shows in the product before a batch. Over a month, track credits per approved ad. My credits guide shows how.
• Stale-data risk. Because it reads your website, a price that changed, or a product that sold out, can end up in a new ad. The person approving must open the live product page.
• Plans. As of November 2026, Higgsfield lists Starter at $19 a month with 270 credits and Plus at $59 with 1,200 credits. Credits from a subscription reset each cycle rather than roll over, so planning a steady weekly batch fits better than occasional large ones. Check the pricing explainer and Higgsfield's pricing page.

The weekly brief and the Monday check

The five-line brief

Write it once, then copy it each week. Change only the lines that need it.

Product: [name, as on the website]
Offer this week: [price or deal, end date]
Fixed claims (allowed list): [the only claims you may use]
Hook angle to test: [one angle, e.g. speed, price, trust]
Banned words: [words you never use]

Pair it with a pre-approved brand kit. Lock the logo, colours, legal line and the allowed claims once. Then each week's output needs only a visual check and a number check.

The Monday check (10 minutes)

1. Open your ad report for the last 7 days.
2. For each ad, look at frequency, CTR and cost per result.
3. Mark an ad for refresh only if frequency and CTR move the wrong way together.
4. Leave healthy ads alone.
5. Fill in the brief for the ones you are refreshing.

Approval and the log sheet

A person approves every new ad before it runs. The approver opens the live product page and checks four things: price, product name, stock status and the end date of any offer. Then they check the logo and the text, reading every word.

Keep a simple log with these columns:

Column	Why it matters
Date	Shows how long each ad ran
Product	Compare products over time
Ad file name	Find the exact image later
Hook angle	Learn which angle wins
Placement	Some shapes behave differently
Spend, CTR, frequency	The signals above
Kept or killed	The decision and the reason

Three weeks of rows will teach you more than any tool's advice. If you also keep each file name in a pattern like product_angle_ratio_v01, you can match the log to the image without guessing. For sizes and safe zones, see resizing ads for all placements.

A 4-week test with pass and fail rules

A plan you can run. The numbers are made up for illustration.

Say you have 8 products and one new ad each per week, which is 8 ads. At 5 minutes of review each, that is 40 minutes of checking a week. Automation can cut the making time, but it does not cut the review time, so budget for it.

1. Week 1: keep your current ads. Record the baseline CTR, frequency and cost per result.
2. Weeks 2 and 3: refresh only ads that show the warning signs. Change one thing (hook, then visual). Log each change.
3. Week 4: compare each refreshed ad to its own baseline.

Pass: at least half the refreshed ads improve CTR or cost per result against their baseline, no ad was rejected by the platform, and no wrong price or product reached a live ad. Fail: any error reaches a live ad, or refreshed ads do no better than leaving them alone. In that case the refresh was not the problem, so look at the offer or the audience.

What to automate and what to leave to a person

Automate	Leave to a person
Making headline and image variants from the brief	New offers and prices
Naming and saving files	Any claim, testimonial or result
Filling the log from the ad report	Regulated products and legal lines
A Monday reminder	Final approval before an ad goes live
Applying the saved brand kit	Anything with a real person's face or voice

Rules that stay the same whatever the tool: AI output may not be unique and Higgsfield's terms do not guarantee it is. You need rights and written consent for any real face or voice. Keep real footage when a claim must be real. Free Higgsfield accounts add a watermark, and the pricing table marks commercial use as not included for Free while the help center says use is not restricted to paid plans, so I would use a paid plan for live ads. Check each platform's current rules on labelling AI content. This is information, not legal advice.

Who should skip this

• You run fewer than 5 ads. Refresh by hand when the numbers say so.
• You run small budgets where frequency never climbs. You may not need to refresh for weeks.
• Your product list changes daily. Weekly briefs do not fit; you need a feed-based tool.
• You want true hands-off automation. A generator plus a person is not that. Look at tools that document scheduling, and read what they mean.

For other tool ideas, read best AI ad generators for ecommerce and how to make ad creative faster.

How I'd approach this

I would not start with a tool. I would set up the five-line brief, the Monday check and the log for two weeks using a spreadsheet and a saved design template. That tells you which ads actually go stale and how long they last. Only then would I add Ads Studio for the products where the log shows you need new versions often.

When you do add it, keep the batch small, read the credit cost, and let a person approve every file. You can look at Higgsfield and its plans before you commit.

This is where a custom build pays off. Blur Studio can build the flow from brief to variants to approval: a form that holds the weekly brief, a step that makes the variants, an approval page where a person ticks the checklist, and a log that fills itself. See how I set that up.