YouTube today published a written account of how its planned video A/B testing tool will work: a test runs, YouTube reports results, and the creator gets seven days to decide which version of the video to keep. The post, on the YouTube Blog, recaps a conversation between the creator strategist Paddy Galloway and Amjad Hanif, YouTube's vice president of creator product, hosted by YouTube's creator liaison Rene Ritchie, in which Hanif said the company built the tests around retention curves rather than click-through rate because "that is the true insight from the A/B testing."
In Short
YouTube has explained in more detail how creators will be able to test different versions of the same video, starting with Shorts, and then choose within seven days which one to keep, guided by how well each held viewers. This matters to creators, and to the brands that sponsor them, because the opening seconds of a video, where sponsor mentions often sit, are the part creators are expected to test most. The same conversation also described tools that show different thumbnails to different viewers, refresh old videos and draft brand pitches with AI, but YouTube still has not said when most of them will arrive.
A conversation from September, written up in November
The source material comes in two parts, seven weeks apart. The first is a video, "Paddy Galloway Dives Deep on Video A/B & Ask Studio", published on September 30, 2026 by the Creator Insider channel together with YouTube Creators. It runs to about 24 minutes in 14 chapters and belongs to the Inside YouTube Podcast series, which listed 49 episodes on November 16. By that date it had 94,429 views, 938 likes and 311 comments. Creator Insider had 937,000 subscribers.
The participants refer several times to announcements made "this week", which places the recording in the week of Made on YouTube, the company's annual creator event, held on September 23. The video description says the three discussed "Video A/B testing, Dynamic Thumbnails, Insights, Research, Ask Studio, and all the new Studio tools announced at #MadeOnYoutube."
The second part is the blog post published today, "Paddy Galloway dives deep with us on Video A/B, Ask Studio, and evolving over time", carried in the Creator and Artist Stories section under Ritchie's byline and listed as a five-minute read. It condenses the conversation and quotes all three participants, with the spoken remarks lightly edited for print.
The video's transcript is generated automatically and garbles several names: Galloway appears as "Patty", Ritchie as "Renee Richie", Hanif as "Amjad Hanife" and "Anad", Ask Studio as "As Studio", the data designer Edward Tufte as "Edward Tuy", and the band Linkin Park as "Lincoln Park". The AI summary YouTube displays under the video repeats the misspellings of the three participants. This article uses the correct names outside quotation marks.
PPC Land set out the full Made on YouTube slate on September 24, and detailed the video testing and dynamic thumbnail demos on September 27. What the conversation and today's post add is the mechanics those first descriptions left open.
Seven days and a predicted winner
When video testing was first shown, its demo did not say how a test ends or what the creator does with the result. The September 27 report noted that the sources disagreed on scope, with the narration covering Shorts or long-form videos and the description pointing to Shorts alone.
The conversation settles both points, at least for the first release. Video A/B testing is introduced in the video as "Starting off with shorts, but this could let you test hooks. It could let you test story structures. It could let you test call to actions." Shorts are therefore where the feature begins, with no date given for longer videos.
The end of a test is described in one sentence: "the test is going to run. It'll give you a predicted winner and then you have seven days to to make your choice." Today's blog post puts the same rule in YouTube's own written voice: "Video A/B testing will run a test, provide results, and give the creator seven days to choose which edit to go with."
That is a different model from YouTube's existing title and thumbnail tests. When YouTube added titles to those tests in December 2025, the variant with the highest watch time per impression was applied automatically once the test ended, and if no winner emerged the first combination uploaded became the default. With video testing, YouTube names a predicted winner, but the creator makes the call.
How many versions can compete? The conversation refers to "the retention for the two the two or three different videos that you're testing", consistent with the limit of three cuts PPC Land reported in September. Each version gets its own curve. "You're going to get both retention curves," the video notes, and the participants suggested the comparison might not produce a clean winner at all. "And the real winner might be this hybrid and that's good," one of them said. "You know there might be two great moments from the two different variants and that's gives you feedback for the next video that you produce."
Retention, not clicks
Why not simply count clicks? Ritchie put the question in terms of thumbnail testing. "When you did thumbnail AB testing, you didn't center on CTR because it could encourage click baiting and then the retention would drop," he said, describing video testing as "giving creators like a fuller picture with with the analysis."
Hanif acknowledged that creators still want the number. "I know creators still ask for CTR on the thumbnail testing," he said. In the blog post's version of his answer: "When you have a sea of metrics, it's easy to get distracted by something that may not actually be in your interest as a creator. That's why we really focused on the retention curves, because that is the true insight from the A/B testing."
The point was taken further in the discussion with an observation about how click-through rate behaves as a video spreads. "Like a CTR is high because no one's watching like as virality increases CTR tends to go down," the video states. A test that shows which version people preferred, by contrast, "is like very actionable information very objective and it's hard to argue with um especially at the scale of YouTube and the viewers that are seeing those different variants so it's more than like a a test with 10 people."
YouTube has made the same choice before. The December 2025 title tests were also set up to optimise for watch time rather than click-through rate, as PPC Land reported at the time. Video testing applies that logic to the content itself.
How the two kinds of test compare
| Title and thumbnail testing | Video A/B testing | |
|---|---|---|
| What is tested | Up to three titles, thumbnails or combinations | Two or three versions of the video itself |
| Formats | Long-form videos, via Studio on desktop | Starting with Shorts |
| How the result is judged | Watch time per impression | Retention curves for each version |
| What happens at the end | Best variant applied automatically; first upload becomes default if no winner | YouTube gives a predicted winner; creator has seven days to choose |
| Test length | Variants shown equally for up to two weeks | Not stated |
| Status | Extended to titles in December 2025 | Announced September 23, 2026; no release date given |
Sources: YouTube (video A/B testing column, from the September 30 video and today's blog post); PPC Land's December 10, 2025 report (title and thumbnail column).
The first 30 seconds
Asked how creators will use the tool, Hanif pointed to the opening. "I think a lot of that energy is going to be on that intro," he said. "I think creators probably that would be the number one thing that I've heard at least from creators is really testing different maybe the first 30 seconds." Different storylines and "maybe a different way of ending" were also on his list, "but I think for the most part it's going to be focused on that intro."
Galloway, who advises large channels, explained why the intro had been out of reach until now. He said he puts "so much emphasis on thumbnails, you know, with our clients because we can retroactively change it." With the video itself, "it's like we've uploaded it, there's very little we can actually do to test." The intro, in his ranking of what moves a video's performance, "is right up there."
His rule for intros is simple. "Your title thumbnail makes a promise. Your intro should deliver on it," he said. Clients often overload the opening, in his account, with "here's everything that's happening. We're going to do this. We're going to do that." He expected some creators to be daunted by the workload at first, and pointed to smaller variations. "I think maybe at first people would be a little bit intimidated with like whoa, I have to like make multiple videos now," he said. "But I think this is where it's important to think about the the subtleties of like testing different intros, testing different CTAs."
Podcasts came up as an obvious case. Galloway said podcasters often wonder: "should we do a a quick compilation of clips at the start or should we just start with the first question? Now we can actually test that." Hanif said a podcaster had told him earlier that they wanted to test "the trailer versus the non-trail intro."
Hanif expected the tool to puncture creator folklore, citing beliefs such as "oh, I think an intro has to be 23 seconds.5." He framed the benefit in time saved: "so much time gets invested in all these myths or things that you thought were made sense and this will help you just really focus on what actually moves the needle."
On-camera presence is one of those assumptions. "We also learned a lot from the early thumbnail testing is sometimes being on camera is really great," Hanif said, "but sometimes really just focusing on what the story is about and not about you, but about what you're talking about also hooks folks, particularly folks that aren't your core audience."
Galloway called himself someone who has "probably studied 10,000 plus retention curves like second by second", looking for patterns such as whether a brighter background holds viewers better than a darker one. "It's beautiful that now we can just do that with certainty," he said.
Three thumbnails, one viewer at a time
The second feature discussed in depth was dynamic thumbnails. Hanif described it plainly: "for dynamic thumbnails, you can upload three three thumbnails and based on what we know about the viewers, we will serve the best thumbnail for that particular viewer." The process is automatic. "You just have to give us three thumbnails."
The blog post puts it as going "beyond regular 'A/B' testing: you provide three thumbnails and they will be served to different audience segments automatically based on which one performs best for that group."
Ritchie offered his own reading. "One for me is segmenting by like new versus regular versus casual audience," he said. "Like maybe for my regular viewers, I'm going to have my face in it. Maybe for new viewers, I'm going to have like the gadget or the device in it because they don't really know or care who I am." Those three groups match the categories YouTube introduced to Studio analytics in July 2025, when it split returning viewers into casual viewers, who watch at least once a month for one to five months of the past year, and regular viewers, who do so for more than six.
Ritchie also pointed to "pocket audiences", niches that respond to one angle of a video, which creators have discovered through A/B tests. Galloway said this explained a pattern he had seen with clients. A thumbnail test would produce a result, and "a week later I'm like I really think the one that narrowly lost is actually might be a bit better with the video that or the audience that's being surfaced this video right now." A rerun could flip the outcome. In the blog's wording of his conclusion, "different audiences sometimes do require a different thumbnail."
Until now, the workaround was guesswork. "Generally speaking, in the first day, we might get a few more core viewers. So, I'm going to try this thumbnail for the first day," Galloway said. "But we all know you're still getting new viewers on your first day." His verdict: "It's always a good sign when you're adding things that like people like me have been trying to manufacture and do behind the scenes for years." Galloway made a similar point on the Made on YouTube post show, which PPC Land covered on October 11.
What neither source explains is how YouTube assigns a viewer to a thumbnail, or whether creators will see results broken down by audience group.
An agent for the back catalogue
The third Studio feature is aimed at older uploads. "Studio will also work behind the scenes for your back catalog," Hanif said, describing an announcement made that week. His example was a video that starts "to find a new audience and getting traction" two years after upload. Ask Studio, he said, will "identify those videos and suggest maybe new thumbnails for them to modernize it to really connect with whatever is happening in the outside world that's given it a new life."
The video description files the chapter under "Back Catalog Optimization"; elsewhere in the conversation it is called "the new background optimization feature". It is described in the video as a tool that "will go through and let you do either run it as a new thumbnail or run it as an AB test and get those answers." The blog post calls it a feature "creators may not know about", in which Ask Studio "will identify older videos that could benefit from a new modernized thumbnail to help them perform their best."
For channels with a deep back catalogue, the appeal is labour. "Of course you could do it on your own but that's a lot of work if you're really focused on on creating your content," Hanif said. Ritchie later grouped it with other agentic features: "we're starting to look at agentic things like the background optimization can do a lot of the heavy lifting for you."
Hanif said each feature produces feedback that shapes the next one, and that thumbnail tests revealed that "creators were spending a lot of time on very minute differences that did not move the needle." Galloway's own example of a myth that testing dispelled was "the creator who always thought like red text was better than white text."
Drafts, stories and brand pitches
The conversation opened with the tools that come before upload. Hanif said YouTube is "investing in pretty much every stage" of creation, from the first idea, where Ask Studio will, in his words, "help you think through ideas based on what we know about your audience", to the storytelling assistant.
"You can put that draft idea, that story line in, and it will give you feedback on the story line, suggest edits to it to make a stronger hook, to capture your audience based on everything studio knows about your viewers and the audience you built," Hanif said.
Ritchie singled out draft mode. "One of my favorite things about draft mode is you don't have to then upload another video and do the title again and you can just replace the draft mode video," he said. Galloway agreed that "having a working canvas instead of just having something you have to constantly replace is going to be pretty huge."
Today's blog post merges the two. It says that "Paddy points out that one of the most useful parts of Ask Studio is the storytelling assistant, which lets you workshop ideas in a draft state before uploading a completed video." In the video, it was Hanif who described the storytelling assistant; Galloway's remark was about draft mode.
Pitches written from public information
For advertisers, the most direct change is a brand pitch tool. "You can ask Studio to help draft a brand pitch and it will take everything it knows publicly about that company, everything it knows about your audience and your catalog of content and put together an amazing pitch," Hanif said. "It's a great starting point for you to then refine and send onto that brand."
PPC Land's September 24 coverage recorded that Ask Studio would help creators draft custom proposals and respond to brand briefs. Hanif's description adds the inputs: public information about the advertiser, plus the creator's audience and catalogue data held by YouTube. He did not say which public sources are read.
Ritchie described a related use he already has. "When I did the video with Amjad, I asked Studio to tell me like what the comment sentiment was," he said. "Brands have asked me that and before it was like, I don't know, go read the comments." Now, he said, he can "put together a very quick, very comprehensive report on like what the what the sentiment was based on like the read that I did or the language that they used or the messaging." That turns comment sentiment, a figure brands ask for in campaign wrap-ups, into an AI-generated summary. No accuracy data was offered.
Hanif tied all of it to a single principle: to take tasks "that are hard, that are repetitive, um that require deep analytics or deep insight and to give you that information so you can actually make the decisions and you stay in control." Ritchie's version was more domestic, asking how YouTube would keep the tools "in service of the creators, letting [Ask Studio] do the dishes while we make the art?"
Ask Studio, a year on
Ritchie opened the episode with an endorsement. "I'm really enjoying Ask Studio. Uh, and I know it's a cliche, but it does 80% of like the busy work for me now, so I can focus on the 20% of the stuff that really matters," he said. He added that he uses it mostly "to roast the ideas that I already have." The 80% figure is his own estimate, and Ritchie is a YouTube employee.
Hanif offered adoption evidence of a similar kind. He said he had met "a group of creators in Wyoming" a year earlier, when "none of them had used" Ask Studio because "it was just coming out", and that "fast forward a year later, all of them are using it." In his words, "in 12 months, it has become part of their daily routine." YouTube introduced Ask Studio at Made on YouTube 2025 on September 16, 2025, which fits the timeline. The company gave no usage figures.
Two product changes closed the conversation. "We haven't had access on mobile for a while. It's been desktop only," Hanif said. "So, that's another exciting announcement um this week, which is that we'll be available on mobile as well." He added that Ask Studio would keep conversation history, "so you can go back to your old chats and pick them up where you left off."
Both have since shipped in some form. Ask Studio reached the Studio app on Android and iPhone on November 13, alongside the new Insights page that replaced Analytics on desktop and the Research tab that replaced Trends. YouTube's Help Center page for Ask Studio states that conversations are kept for up to 30 days. Earlier in November, YouTube also let creators escalate Ask Studio questions to live human support agents, within limits.
Hanif's account of Insights in today's post echoes that rollout: "What we heard time and again from creators was it's a lot to interpret, it's great if you have a team of analysts, so the spirit behind the work on the new insights is to bring out what's important for you and give you a chance to ask Studio to explain these things and interpret them." In the video he added: "You'll always have the raw data."
Galloway's method
Asked what he wants surfaced, Galloway started with the most basic number. "Views are a metric," he said, and "all analysis for me starts with the view count." Retention curves come next, as "one of the easiest ways to kind of speedrun like learning what your audience likes or doesn't like."
He gave one client example, a large sports channel whose audience, he said, "is basically allergic to like scripted kind of skits". Competitors were using comedy skits, but when the channel tried them, "the audience just consistently just dropped off." The lesson went beyond one video. When the channel leaned "more into like authentic ideas, more personal stories with our video ideas, the channel did better."
He described the rest as "this kind of four or fivestep analysis": views, then the retention curve, then checking whether other metrics such as click-through rate or average view duration explain anything, and finally pulling it together into "an action plan." His standard question to creators at events is "What did you learn from your last video?" Too often, he said, the answer is that they no longer know, because "people get so lost in the numbers."
On outliers, which now feature in Studio's Research tab, Galloway was enthusiastic but cautious. Looking at "what is overperformed for your channel and other channels" is, he said, "one of the most effective ways to learn YouTube and content strategy in general." He traced the habit to tracking the daily views of Linkin Park videos at age 12. But he warned that "there can be people that lean into it too much where you're just constantly chasing what's worked for other people."
His alternative is a funnel with three inputs: internal sources (what has worked on the creator's own channel), external sources such as outliers ("It's never can I repeat this? It's can I like put my own spin on it?"), and original ideas that neither the creator nor anyone else has tried.
Hanif framed the Research tab in similar terms. "It's not easy for every creator to think about that," he said of external trends, "the idea behind the research tab is to just help you broaden your horizon a bit, especially if you're just starting out."
What does Galloway expect from all this? "I'm looking forward to less people having to like message me being like, 'What does this mean? I have no idea,'" he said, adding that the changes should help most "for smaller creators."
How the features work, according to YouTube
According to the video and today's blog post, video A/B testing follows five steps:
- The creator prepares two or three versions of the same video. The first release covers Shorts.
- YouTube runs a test on viewers.
- YouTube provides results, including a predicted winner.
- The results show a retention curve for each version, rather than click-through rate.
- The creator has seven days to choose which version to keep.
Dynamic thumbnails, as described by Hanif, work in two steps:
- The creator uploads three thumbnails for a video.
- YouTube automatically serves each viewer the thumbnail it judges best for that viewer, based on what it knows about them.
Background optimization of the back catalogue, as described in the video:
- Ask Studio identifies older videos that are finding a new audience.
- It suggests new thumbnails for them.
- The creator can apply a suggestion as a new thumbnail or run it as an A/B test.
The tools at a glance
| Feature | What the sources say | Status |
|---|---|---|
| Video A/B testing | Two or three versions; predicted winner; retention curves; seven days to choose; starts with Shorts | Announced September 23, 2026; no release date |
| Dynamic thumbnails | Three thumbnails; best one served automatically per viewer | Announced September 23, 2026; no release date |
| Background optimization | Ask Studio flags older videos gaining traction and suggests new thumbnails, applied directly or as an A/B test | Announced September 23, 2026; no release date |
| Storytelling assistant | Feedback and suggested edits on a draft storyline, including a stronger hook | No date given |
| Draft mode | Replace a draft video without a new upload or retyping the title | No date given |
| Brand pitch drafting | Uses public information about the brand, plus audience and catalogue data | No date given |
| Ask Studio on mobile | Previously desktop only | Available from November 13, 2026, per PPC Land |
| Conversation history | Return to earlier chats | Help Center states conversations are kept up to 30 days, per PPC Land |
Why this matters for marketers
Sponsor reads often sit near the start of a video, which is the section Hanif expects creators to test most. Once versions of a Short can compete, a sponsor's mention may appear differently, or not at all, in the variants viewers see during a test. Neither source says whether brands are consulted, whether reporting separates impressions by variant, or what happens to the losing version's views. Those questions sit alongside the ones dynamic thumbnails already raise about brand approval of packaging, which PPC Land set out in its October 11 report on the post show.
The retention-first design also changes what creators will optimise for. YouTube's choice to steer tests away from click-through rate reflects a judgment that clicks can mislead. For advertisers buying creator content, it means the platform's own testing tools will reward videos that hold attention rather than those that win the click.
The brand pitch tool reverses the usual flow of a deal. Brands that receive inbound proposals from creators may increasingly be reading drafts assembled by Ask Studio from public information about them and from YouTube's data on the creator's audience. Comment sentiment reports produced the same way may reach brand teams as part of campaign reporting.
What the sources do not say
Neither the video nor the blog post gives release dates or eligibility rules for video A/B testing, dynamic thumbnails, background optimization or brand pitch drafting. They do not say when video testing will extend beyond Shorts, how long a test runs before results arrive, what happens if a creator does not choose within seven days, how viewers are assigned to a thumbnail, or which public sources the pitch tool reads.
Key facts
- Blog post: "Paddy Galloway dives deep with us on Video A/B, Ask Studio, and evolving over time", by Rene Ritchie, YouTube Creator Liaison, YouTube Blog (Creator and Artist Stories), November 16, 2026, https://blog.youtube/creator-and-artist-stories/creator-insider-paddy-galloway-youtube-ab-testing-ask-studio/
- Video: "Paddy Galloway Dives Deep on Video A/B & Ask Studio", Creator Insider and YouTube Creators, published September 30, 2026, https://www.youtube.com/watch?v=j7GtybYc6gk. About 24 minutes, 14 chapters, part of the Inside YouTube Podcast series.
- Participants: Paddy Galloway, creator strategist; Amjad Hanif, YouTube VP of Creator Product; Rene Ritchie, YouTube Creator Liaison.
- Features discussed: Video A/B testing; Dynamic Thumbnails; Back Catalog Optimization (also called background optimization); storytelling assistant; draft mode; brand pitch drafting in Ask Studio; Insights; Research tab; Ask Studio on mobile.
- Video A/B testing rule, per the blog post: "Video A/B testing will run a test, provide results, and give the creator seven days to choose which edit to go with."
- Scope, per the video: "Starting off with shorts."
- Dynamic thumbnails, per Hanif: three thumbnails, with the best one served per viewer automatically.
- Release dates: none given in either source for the testing, thumbnail, back catalogue or pitch features.
- Announcement event: Made on YouTube, September 23, 2026.
Timeline
- April 10, 2025 - A YouTube Studio product director describes Studio as a creative partner and hints at title A/B testing
- July 24, 2025 - YouTube splits returning viewers into casual and regular groups in Studio analytics
- September 16, 2025 - Made on YouTube 2025 introduces Ask Studio and three title and thumbnail combinations per test
- December 9, 2025 - YouTube adds titles to A/B testing, choosing winners on watch time per impression rather than click-through rate
- July 25, 2026 - Custom Shorts thumbnails roll out without A/B testing
- September 23, 2026 - Made on YouTube 2026 announces video testing, dynamic thumbnails, back catalogue optimization, brand pitch help and Ask Studio on mobile
- September 25, 2026 - Creator Insider demonstrates testing of up to three cuts of one video and dynamic thumbnails
- September 30, 2026 - Creator Insider and YouTube Creators publish "Paddy Galloway Dives Deep on Video A/B & Ask Studio"
- September 30, 2026 - The Made on YouTube post show, with Galloway among the guests, is published
- November 6, 2026 - Ask Studio gains escalation to live human support agents for eligible creators
- November 13, 2026 - Ask Studio reaches the Studio mobile app; Insights and Research arrive on desktop
- November 16, 2026 - The YouTube Blog publishes Ritchie's written account, stating the seven-day choice window
Related PPC Land coverage
- YouTube to test up to 3 versions of one video on real viewers, Shorts too - the first detailed look at video testing and dynamic thumbnails, and the gaps they left.
- YouTube Studio's numbers learn to talk as Analytics becomes Insights - the November 13 rollout of Insights, Research and Ask Studio on mobile.
- YouTube creators in 35 markets to earn shopping commissions by year-end - the full Made on YouTube 2026 slate, including Ask Studio's brand pitch help.
- Neal Mohan bets YouTube's future on brand deals and 35-country affiliates - the post show in which Galloway, Marques Brownlee and others weighed the testing tools.
- YouTube expands A/B testing to include titles alongside thumbnails - how existing tests pick winners on watch time per impression.
- YouTube ends 2-year wait for Shorts thumbnails but blocks A/B testing - custom Shorts thumbnails arrived without testing in July 2026.
- YouTube introduces refined viewer analytics with casual and regular metrics - the new, casual and regular viewer definitions behind audience segmentation in Studio.
- YouTube Studio now suggests products for affiliate creators to tag - the shopping shelf and Ask Studio's link to human support.
- YouTube announces major creator tools with Veo 3 integration and $100 billion payout milestone - the September 2025 event where Ask Studio was introduced.
Summary
Who: YouTube, through its VP of Creator Product Amjad Hanif and Creator Liaison Rene Ritchie, in conversation with the creator strategist Paddy Galloway.
What: A written account of how YouTube's planned video A/B testing will work: two or three versions of a video, starting with Shorts, judged on retention curves rather than click-through rate, with a predicted winner and seven days for the creator to choose. The conversation also described dynamic thumbnails served per viewer, background optimization of older videos, a storytelling assistant, draft mode, AI-drafted brand pitches, and Ask Studio on mobile with conversation history.
When: The blog post was published today, November 16, 2026. The video it summarises was published on September 30, 2026, after the announcements at Made on YouTube on September 23.
Where: On the YouTube Blog and the Creator Insider YouTube channel, about features in YouTube Studio and its Ask Studio assistant.
Why: Creators can change a thumbnail after upload but not the video itself; video testing extends testing to intros and story structure, where sponsor messages often sit. YouTube built the tests on retention rather than click-through rate, which its creator liaison said could encourage clickbait, and the brand pitch tool will put AI-drafted proposals in front of advertisers. Release dates for most of the features have not been given.
Discussion