Production guide

The ad design checklist: 24 checks before an ad goes live

Four tests every ad should pass at phone size, turned into 24 quick checks with the reason behind each one, plus a worked review of three ads from one batch.

Guide · No. 23

Reading time
10 minutes
Updated
Topic
Production

The short version

  • Review ads on a phone, in the real preview, never only on a big monitor. That's where people will see them.
  • Four tests catch most problems: the product reads as a thumbnail, the words land in one glance, nothing sits under the interface, and every claim can be proved.
  • Measure text contrast instead of judging it by eye. The WCAG bar of 4.5:1 for small text and 3:1 for large text is a fair target.
  • Run a second pass on the whole batch: one visual family, genuinely different ideas and a clear label on what each ad tests.
  • Automated checks catch spelling and product drift. Offers, prices and the truth of a claim still need a person.

The short answer

Before an ad goes live, check four things at phone size: people can tell what you sell from a thumbnail, the words on the image land in one glance, nothing important sits where the interface or a crop will cover it, and every claim is one you can prove. The checklist below turns those four tests into 24 quick checks in six groups, with the reason for each and a worked review of three ads. It works on anything: AI-made ads, a designer's files or your own photos.

How to run the checklist

  • On a phone, in the real preview. A big monitor flatters every ad. Open the preview for each placement in Ads Manager, or send the image to your phone and look at it in your hand. Before upload, the free ad mockup generator shows your image and text in a generic feed, Stories or Reels frame.
  • In two passes. The first pass is ad by ad. The second is the batch as a whole: does it look like one brand, and does each ad test a different idea?
  • With fresh eyes. Whoever made the ad sees what they meant to make. A second person sees what is there. If you work alone, wait an hour before the second look.
  • With a rule for failures. A failed product or claims check means fix it or drop the ad. A failed contrast or edge check can often be fixed by changing one element.

1. The product

  • Recognisable as a thumbnail. Shrink the ad to a quarter of its size. If you can't say what kind of product it is, the image is asking the headline to do all the work.
  • The real product, not a lookalike. Label text, cap colour, shape and proportions match your packshot.
  • Whole and uncut. No lids, handles, straps, nibs or shoe toes lost off the edge of the frame.
  • Honest scale. A 200 g jar shouldn't read as a family tub. When size matters, put a hand, a spoon or a mug in the frame.

The product group comes first because every later check assumes it passes. The guide to product photos for AI ads covers shooting a clean starting photo on a phone. In BulkAds, every ad made from your product photo is compared with that photo before you see it, and drifted ads are flagged. Glance at the label anyway: a flag says something changed, your eyes say whether a customer would notice.

2. The words on the image

  • One line that makes sense alone. A few words, one idea. If you need a second idea, it belongs in the primary text.
  • Spelt right, letter by letter. Read each word on its own at phone size, including the small ones. AI image models still garble short words and numbers now and then.
  • The same idea as the copy. The line on the image, the headline field and the first line of primary text should push one angle, not three.
  • Prices and terms match the store. The price, discount, end date and free shipping threshold say exactly what the landing page says on launch day.

As of October 2026, Meta's photo ad format page (opens in a new tab) advises keeping text on images to the essentials and asking people to look at only one thing, and suggests a carousel or a video when one image is trying to show too much. The on-image line is a sign on a shop window, not a product description.

The price check is the easiest to skip because it feels like admin. An ad that says one deal and a page that shows another makes people feel tricked at the moment they were ready to buy.

3. Contrast and legibility

  • Text contrast passes a real test. Measure it with any free contrast checker using the two hex codes. Don't trust your eyes on a bright screen.
  • Text sits on a calm area. Words never run across a patterned rug, a face or the product. Move the line or give it a solid panel.
  • The order survives grayscale. Turn the ad black and white. The product and the line should still be the first two things you notice.
  • The smallest text reads without zooming. If the small print can't be read at arm's length, cut it or move it into the copy.

For the contrast number, borrow the bar from web accessibility. The W3C's WCAG 2.2 contrast criterion (opens in a new tab) (checked October 2026) asks for a ratio of at least 4.5:1 for normal text and 3:1 for large text, which it defines as at least 18 point, or 14 point bold. Text that is part of a logo or brand name has no requirement. It was written for web pages, not ads, but a phone outdoors is a hard place to read, so it's a fair target.

Points don't map neatly onto an ad file. In CSS, 18 point is 24 pixels. A 1080-pixel-wide feed image shown full width on a phone about 390 pixels wide shrinks to roughly a third, so a line needs type around 66 pixels high in the file to count as large, or about 52 pixels if it's bold. Anything smaller should reach 4.5:1.

Brand palettes fail more often than you'd expect, usually with white type on a mid-tone brand colour. By the WCAG formula, white on a bright tomato red such as #E8452C comes out at about 3.95:1. That clears the bar for a big headline and fails it for the small line underneath. A shade in between, #D9402A, lands at 4.45:1, just short, which is why measuring beats guessing. Darkening the red to #C63A22 lifts white type to about 5.2:1, a change few customers would notice. Put the darker shade in your brand kit so every future ad uses it.

4. Edges, crops and the interface

  • Composed for its shape. A square, a 4:5 and a 9:16 version are three compositions, not one image stretched three ways. The aspect ratio entry explains the shapes.
  • Vertical versions stay clear of the interface. Keep words, logos and the product out of the zones the profile name, captions and buttons cover.
  • Margins for automatic expansion. Leave quiet background around the subject, so any extended edges never meet the product or the text.
  • Video works with sound and without it. The voice and music should help, and the captions should carry the story on their own.

For the vertical check, Meta's Ads Guide entry for Instagram Stories images (opens in a new tab) suggests, as of October 2026, leaving roughly 14% of the top, 35% of the bottom and 6% of each side free of text, logos and other key elements. The free safe zone checker draws approximate zones over your own ad.

Expansion matters because Meta may extend your image for you. Its Reels ads page (opens in a new tab) says, as of October 2026, that image expansion uses generative AI to extend a creative to the 9:16 Reels frame. You don't choose what gets painted in, so give it plain background and check the vertical preview. The Advantage+ creative entry covers Meta's other automatic edits.

The same page says Reels default to sound on unless the viewer has turned it off, but some people still watch muted. The guides to captions for video ads and the video ad first frame cover both halves.

5. Claims and policy

  • Every claim can be proved. Numbers, ratings, "lab tested", "No. 1". If you couldn't show the evidence on request, take it out.
  • No lines about personal attributes. Talk about the product, not about what the viewer is or has.
  • No fake interface. No drawn play buttons, notification badges, checkboxes or "tap here" buttons on a still image.
  • Bodies and before-and-afters handled with care. Check the rules for your category before any transformation shot goes near an ad.

Meta's personal attributes policy (opens in a new tab) (checked October 2026) says ads must not assert or imply personal attributes, and its examples of what's not allowed include direct questions about a person's health, finances, religion or age. The design fix is usually small: turn the question into a statement about the product. "Dry skin?" becomes "For skin that feels tight by noon."

For transformations, Meta's health and wellness policy (opens in a new tab) (checked October 2026) lets general cosmetic products show before-and-after transformations only to audiences aged 18 and over, doesn't allow weight loss ads with a close-up of pinched body fat, and rules out statements of inferiority about physical appearance. Its Advertising Standards (opens in a new tab) also say video ads must not use overly disruptive tactics such as flashing screens.

A drawn play button on a still promises an interaction that isn't there, and whoever taps it gets a landing page instead. While you're here, check that no real competitor's pack, logo or name appears, and no person you don't have the rights to show.

6. Brand and the batch

  • The brand shows without the logo. Cover the logo with your thumb. If the ad could belong to anyone, the colours, type and pack aren't doing their job.
  • One visual family. The same palette, type and product treatment across the batch, so the twentieth ad still looks like the first.
  • Genuinely different ideas. Lay the batch out as a grid. Ads that differ only in background colour count as one creative concept, not several.
  • Each ad labelled with what it tests. The angle, the hook type and the format go in the ad name, so the results mean something later.

Meta's photo ad format page also suggests that your images share a consistent theme and tie together visually. That covers the first two checks. The last two are about learning. Twelve near-identical ads can't tell you which idea sells, and the guide to concept vs iteration testing explains how to spread a batch across ideas before you spend it on variations. Ads launched through BulkAds are named by angle and ID for you, so the label check matters most for ads made elsewhere.

A worked review of three ads from one batch

Take Fyrlo, a fictional chilli crisp sold in a 200 g jar at $4.65. Its first BulkAds batch is twelve ads in 4:5, each with its own angle, made from one packshot. At 2 credits per image ad, the batch uses 24 credits. Nine ads pass both passes. Three don't, and between them they show which checks a machine catches and which only a person can.

Ad The check it failed Who caught it The fix
Bold claim: "Crunch in every spoon." in white on tomato red Contrast. The big line clears 3:1, the small "200 g, small batch" line underneath falls short of 4.5:1 The reviewer, with a contrast checker Set the brand red to #C63A22 in the brand kit, then re-render from the same concept
Lifestyle: a bowl of noodles with the jar on a shelf behind it Thumbnail. At a quarter size the jar is a red dot and the label can't be read The reviewer, at quarter size Remake it with the jar in front of the bowl
Offer: "Buy 2, get 1 free" with a drawn "Claim now" button Fake interface, and the store actually runs "3 jars for $11" Pre-flight flagged the button. Only the reviewer knew the offer Remake it with no button and the live offer

None of the three was a bad idea. Each failed on something a customer would notice and its maker might not. The first fix is a re-render. The other two are new ads, at 2 credits each. Had the offer ad gone live, every click would have landed on a different deal from the one the ad promised.

What BulkAds checks for you and what stays with you

Automated review is good at mechanical checks, not at knowing what your store says today.

Check What BulkAds does You
Product matches your photo Scored against your packshot, drifted ads flagged A last look at the label
Spelling on the image Checked word by word Names, prices and dates
Brand fit and visual quality Scored on palette, style, composition, artefacts and whether the product is the hero Whether it feels like you
Policy risk Flags before-and-after body shots, medical claims and fake buttons The final call on borderline claims
Claim is true Copy is written only from the proof points in your brand kit Keeping the brand kit honest
Offer matches the store Not checked Every time
Interface and crops Not checked Preview, plus the safe zone checker
Variety across the batch Each ad takes its own angle A quick look at the grid

Ads that fail are marked. You can rewrite the headline, primary text and description in place, or re-render a flagged image from the same concept, without starting a new batch. The bulk ad launcher then puts the clean batch into your own Meta ad account paused, so there's one more chance to look at the real previews before anything spends.

When breaking a rule is the point

Some of the best ads break a check on purpose. A lo-fi creative may put the product small in a messy kitchen because that's what a customer's photo looks like, and a notes-app ad may show no product at all.

That's fine when the break is a decision. Say so in the ad's name ("product small on purpose") and launch it next to a version that follows the rule. The claims group is the exception. A false claim or a fake button doesn't make an ad more native, only more likely to be rejected or to disappoint the people who click.

Questions people ask

What should an ad design checklist include?

Checks in six areas: the product (recognisable, accurate, uncut, honest scale), the words on the image (one line, spelt right, matching the copy and the store), contrast and legibility, crops and the interface, claims and policy, and the batch as a whole (one visual family, different ideas, clear labels).

How much text should go on an ad image?

As little as the idea needs, usually one short line. As of October 2026, Meta's photo ad format page advises keeping text on images to the essentials and asking people to look at one thing. Detail, terms and second ideas belong in the primary text.

What contrast ratio should ad text have?

A fair target is the web accessibility bar from WCAG 2.2: at least 4.5:1 for normal text and 3:1 for large text (18 point, or 14 point bold). Measure it with a contrast checker using the two hex codes rather than judging by eye.

Does BulkAds check ads before I launch them?

Yes. Every ad gets a pre-flight check that scores product accuracy against your photo, on-image spelling, brand fit and visual quality, and flags likely policy risks such as before-and-after body shots, medical claims and fake buttons. Offers, prices and safe zones still need your own look.

Should I check every ad in a batch or just a few?

Every ad on the first pass, because each one can fail in its own way. The second pass looks at the batch as a whole, which takes a minute: lay the ads out as a grid and check that they share one look and test different ideas.

Keep going

Written by the BulkAds team. Published , last checked . Meta changes its tools often — if something here no longer matches Ads Manager, tell us at contact@bulkadsai.com.

BulkAds AI is an independent product and is not affiliated with, endorsed or sponsored by Meta Platforms, Inc. Facebook and Instagram are trademarks of Meta Platforms, Inc.

Your next fifty ads are one link away.

Paste your website, meet your brand kit and see your first batch in minutes. 7 days for $9.99 — cancel anytime.

yourbrand.comMake my ads

See pricingExplore featuresBrowse the library