What's changing

A standard ChatGPT ad takes one image, set through a single file_id in the API. The new format, which OpenAI's documentation calls ads with multiple images, lets one ad hold a row of pictures instead. Based on the help centre and the API reference, the rules are:

• Images: four to six distinct square images, each at a 1:1 ratio. The API takes them as an ordered list, so the order you upload is the order the ad uses.
• Copy: one title, one description and one destination URL. The API reference says the title, body and destination URL "apply to every image", so you can't send each picture to a different product page.
• Call to action: OpenAI's help centre says "Learn more" is currently the only option for this format.
• No cropping: the API rejects the image crop setting on multi-image ads, along with the price and business agent fields. Eccel called the missing crop tool a nuisance. In practice your images need to be square before you upload them.
• Campaigns: the format works in standard Reach, Clicks and Conversions campaigns. The API reference says multi-image ads use fixed bidding unless an account is also enabled for automatic bidding with this format.
• Access: the account has to be enabled for multi-image ads. If you don't see the option, your account may not have it yet.

Two limits on editing are worth knowing. You can't turn an existing single-image ad into a multi-image one: OpenAI's API reference says to create a new ad instead. And for a multi-image ad, sending a new list of images replaces the whole set rather than adding to it.

The help centre says these ads show below a ChatGPT response.

Why it matters for advertisers

ChatGPT ads have so far given you one picture to make the case. Four to six images give retailers room to show a range, a product from several angles, or a before and after. Service businesses can use the extra frames to walk through steps. It also brings ChatGPT closer to the carousel-style formats media buyers already run on Meta and in Google's Demand Gen.

The restrictions shape how useful it is. A single landing page for every image means the format suits a category or collection page better than a set of separate products. The fixed "Learn more" button limits how hard the ad can push for a sale. And because the price field isn't allowed, you can't show prices through the ad's own fields even though the API accepts a price on other ad types.

What to do now

• Check Ads Manager for the multi-image option when you create a new ad in a Reach, Clicks or Conversions campaign. If it isn't there, your account likely isn't enabled yet.
• Prepare square assets at 1:1 before you start. With no crop tool, a landscape product shot will need resizing in your own design tool first.
• Point the ad at a page that covers every image, such as a collection or category page, since all images share one URL.
• Test it against your single-image ads in the same ad group rather than replacing them. Compare click-through and conversion rate before moving budget.
• API users: upload each image first, then pass the file IDs in the creative's file_ids list, and plan to create new ads rather than editing existing single-image ones.

Until now, advertisers who wanted these numbers had to build them as custom metrics, with formulas that differ from one agency or tool to the next.

What's changing

Hook rate and hold rate are the two most common ways to judge a video ad's creative before looking at cost or conversions:

• Hook rate measures how many people stopped scrolling once the video started. It is usually calculated as 3-second video plays divided by impressions.
• Hold rate measures how many of those people kept watching. The usual version divides ThruPlays, which Meta counts when a video plays to completion or for at least 15 seconds, by 3-second plays. Some teams divide by impressions instead.

With native columns, these ratios can sit in a saved column preset next to spend and results, without anyone maintaining custom formulas. Meta has not yet documented which formulas its own columns use, so check the column descriptions in your account before comparing them with numbers you calculated yourself.

Why it matters for advertisers

The main benefit is consistency. Because every team has calculated these metrics its own way, a "30% hook rate" from one agency could not be compared with the same figure from another. A single definition from Meta gives advertisers, clients and creative teams a shared number.

There is a catch for anyone with historical reporting. If Meta's formula differs from the one in your custom metric, the native column and your old numbers will not match. Benchmarks built on your own formula may need to be rebuilt before you compare old and new creative.

The change also signals where Meta wants advertisers to look. With Advantage+ automating targeting and placements, the creative is one of the few levers advertisers fully control, and these columns put creative quality on the main reporting screen.

What to do now

• Check whether your accounts have the columns. Look under Columns in Ads Manager, in the video engagement metrics. If they are missing, the rollout may not have reached you yet.
• Compare Meta's numbers with your custom metrics on a few recent ads before switching reports over, and note any gap.
• Use the two rates together to find where a video loses people, as Eccel lays out in his post. A low hook rate points to the first three seconds. A strong hook with a weak hold points to the middle of the video. Strong rates with few conversions point to the offer or the landing page, not the creative.
• Add both columns to your creative testing preset, so every new video is judged on the same metrics from day one.

The same post confirms that multiturn ads, which can appear when a user asks a follow-up question inside an AI Mode conversation, are launching in the US. Google did not give a date.

What's changing

For most accounts, nothing needs switching on. According to Google's help page on ads in AI experiences, Search, Shopping and Performance Max campaigns are eligible automatically. There are no new campaigns to build and no separate budgets. Ads enter the normal auction when Google judges them relevant to the conversation, and they carry a "Sponsored" label.

The help page sets out a few details the announcement leaves out:

• Formats: text ads and Shopping ads. Local inventory ads and vehicle ads are in testing.
• Matching: broad match, AI Max for Search, Shopping and Performance Max can match both the user's question and the context of the AI response. Exact and phrase match keywords only match queries with direct commercial intent.
• Markets: the page describes availability as English in the US and select English-speaking markets. Pakistan is listed as text ads only.
• Control: you cannot target AI Mode on its own, and you cannot opt out of it separately from Search. Negative keywords and brand exclusions still apply.
• Reporting: AI Mode impressions, clicks, conversions and cost are folded into standard campaign reports, and the triggering queries appear in aggregate in the search terms report. There is no dedicated AI Mode segment.

Google backs the rollout with two data points. AI Mode queries are on average three times longer than traditional searches, from its own US report in May 2026. And click-through rates for sponsored and organic product listings inside AI Mode are "already similar", based on internal US data from late August to early September.

Why it matters for advertisers

In ten more countries, a share of your Search, Shopping and PMax spend can now go to placements you cannot isolate, bid on separately or report on separately. If performance shifts in these markets over the coming weeks, AI Mode is one possible cause, and the reports will not show you directly how much it contributed.

Longer, conversational queries also favour accounts that run broad match or AI Max, since those are the match types that can respond to the AI answer's context. Accounts built mostly on exact and phrase match will only show for queries with clear commercial intent, which probably means less AI Mode volume.

The Q4 markets carry the most uncertainty. Brazil and Mexico are not English-speaking, while the help page still describes ads in AI Mode as English-only. Google has not said whether those launches will come with Portuguese and Spanish support.

What to do now

• Annotate the rollout date in accounts that run in the 10 markets, so later swings in CPC, CTR or conversion rate can be checked against it.
• Read the search terms report more often. Look for longer, question-style queries that suggest AI Mode traffic, and add negatives where intent is off.
• Check brand exclusions and negative lists before more spend flows into these placements, because they are the only controls Google offers here.
• Tidy your Merchant Center feed. Shopping ads are one of the two formats that serve, and Google's guidance points to complete, current product data.
• For UK, Brazil and Mexico accounts, settle a baseline of current performance now, so you have something to compare against when the Q4 launch arrives.