Types of Facebook Ads: Formats Explained and Which to Test First

Facebook ads come in ten commonly listed formats: image, video, carousel, collection, slideshow, Stories, Reels, lead, Advantage+ catalog and Instant Experience. A format is how the ad is built; a placement is where it appears. Stories and Reels are placement-led: they are named for where they run, and the creative inside is usually an image or a video.

Several labels moved in Ads Manager in 2026. In many accounts, Flexible no longer appears as a separate format in Ad setup, and Collection can be selected in the Ad creative step under Format display options. Meta also reduced its campaign objectives to six. Menus differ by account, so check your own setup screen before following an older tutorial.

As of November 2026, guides that rank for this search list between 7 and 13 formats: AdEspresso states 7, Sovran 10, Ryze 11, BigCommerce 12 and Young Urban Project 13, while AdAdvisor states that there is no single correct count. The counts differ mainly because some lists treat placements such as Reels and Stories, or messaging ads, as formats. Ten is a workable list for planning tests, not an official count. For each format in use, see Facebook ad examples.

Facebook ad types (formats), placements and objectives are three separate choices. The objective is the result you ask Meta to optimize for. Meta now offers six: Awareness, Traffic, Engagement, Leads, App promotion and Sales. The format is how the ad is built, such as video or carousel. The placement is where people see it, such as Feed, Stories or Reels.

One ad uses all three: a video (format) optimized for sales (objective) and shown in Reels (placement). Newcomers mix them up because the names overlap. "Reels ads" names a place, while "video ads" names a build. Only the objective, such as Leads, names the result you want.

This matters for testing. If you change the format and the placement together, a better result could come from either, and you cannot tell which. Fix the objective and placement first, then vary the format. A format test is only readable when the format is the only thing that moved.

An image ad is one still picture with text and a button, the simplest Facebook ad format to produce. Test it first when you want a clean control (the baseline ad you compare new formats against), because a still isolates your offer and angle from any effect of motion.

Competitors show you how they use it. In the Ad Library (Meta's public database of ads currently running), note whether their longest-running stills show the product alone, a person using it, or text on a plain background. Those are three different visual bets.

The hypothesis an image tests is: this offer works without motion. If a still holds its own against video, you have saved production cost on every later test. For image against video, see static vs video ads. A customer review set on a still is a common angle, covered in Facebook testimonial ads.

A video ad is one clip with text and a button, delivered in Feed, Stories or Reels. Video earns its cost when your product needs motion to be understood: how it is used, how it feels, or what happens next.

Check whether competitors repeat the same opening seconds across several videos. A repeated hook is a pattern worth testing; a one-off is noise. Stories and Reels are full-screen placements, but the clip inside is still a video ad, so you can test the same video across placements later.

The hypothesis a video tests is: the buyer needs to see this working. Reels placement is covered in Facebook Reels ads. Creator-style video is covered in UGC ads, and ads that run through a creator's account are covered in Meta partnership ads.

A carousel is a swipeable row of cards, each with its own image or video and link. A collection pairs a cover image or video with product tiles that open a full-screen page, and in many accounts it now sits under Format display options in the Ad creative step.

Look at whether competitors show one hero product or several. If the same brand keeps running multi-card ads for weeks, range is probably part of its pitch.

A carousel tests whether showing range, steps or variants beats one message. A collection tests whether browsing helps the buyer decide. Both need real choice to show; with one product, a carousel is often just a slideshow of the same idea. Ads built from a product catalog are covered in the dynamic ads section below.

A lead ad opens a form inside the app to collect contact details. An Instant Experience ad opens a full-screen mobile page after a tap, built from images, video and text. Both keep the buyer inside Meta instead of sending them to your website.

A lead ad tests whether asking for details in the app beats a landing page. Use it only when leads are your objective. An Instant Experience tests whether a faster, more visual page beats your website for a first visit. For a store selling a single product, a plain product page is usually the stronger control.

What are dynamic ads (Advantage+ catalog) and flexible ads?

A format only fixes the container. The offer, angle and opening seconds inside it decide whether anyone stops, and those are what deserve the most testing. Format is one variable among several, not the answer.

Meta's third-quarter 2026 results, filed on October 21, 2026, report that ad impressions across its Family of Apps (Facebook, Instagram, Messenger and WhatsApp) rose 10% year over year and the average price per ad rose 14% (Exhibit 99.1). Your own cost depends on your auction, audience and creative.

Our view: with more impressions being sold and a higher average price per ad across the platform, testing more creative ideas, and choosing what to test from data instead of a format table, is the sounder use of a test budget.

Meta's engineers wrote in December 2024 that a planned change to its Andromeda retrieval model aims to deliver more diverse ad candidates, and that "Increased ad diversity can improve people's experience with ads and drive better advertiser outcomes." (Engineering at Meta). That describes Meta's own system, not a promise to any one advertiser.

So the useful question is not which format wins. It is which format your evidence says to test first.

Test first the format your competitors keep running over several weeks that you have never tried, once you have fixed one offer and angle. With no signal, start with a static image as the control. What competitors repeat is a signal about what they run, not proof that it performs.

Use this six-step card. Each step has a pass and a fail.

What is a sensible default test order when you have no signal?

A default order is a starting point for when you have no competitor or account signal, not a rule. Start with a static image control, then a short video, then a carousel or Advantage+ catalog ad if you have a product feed. Add lead ads or Instant Experience only if your objective needs them.

Hold constant across the tests: offer, angle, audience, budget, placements, landing page and run window. Once one ad earns its place, keep it live and line up its replacement, as evergreen Facebook ads explains.

Deepsolv is a creative intelligence platform for brands and agencies advertising on Meta. We turn competitor moves, customer signals and past ad learnings into a ranked weekly test plan. That puts the six steps above on a weekly rhythm, so the format you test first comes from evidence instead of a table.

Pick one offer and angle, then choose your first format by evidence, not by a list. If a competitor has kept a format live for weeks and you have never run it, test that. If you already have results, build on the format with real volume behind it. If you have neither, run a static control, then a short video, and change one thing at a time. That order keeps every result readable, and each format earns its place with evidence. Write the hypothesis and the stop rule before you launch, because a format test you cannot read is spending without learning. If keeping that competitor log by hand each week is the bottleneck, book a demo and see how Deepsolv turns it into a ranked test plan.