At Made on YouTube on September 23, 2026, YouTube announced Custom Feeds, which let viewers type a plain-language prompt and get a dedicated homepage tab that keeps refilling itself, plus A/B testing of up to three full video cuts inside YouTube Studio. Custom Feeds is coming to US viewers on web, mobile and TV. YouTube only said “soon,” but Android Headlines reported a launch next month, which lands right as Michigan’s heating, winterization and year-end budget season starts.
This playbook is for the small Michigan team that runs a YouTube channel alongside everything else: the Livonia HVAC contractor, the Sterling Heights tooling shop, the Traverse City marina. The question is how to make a channel that a prompt can find, not just a subscriber.
What are YouTube Custom Feeds?
Custom Feeds are homepage tabs a viewer creates by describing what they want, like “new DIY creators” or “evening wind down with long documentaries.” Each one updates automatically with matching videos and channels. A viewer can save several, each tied to a mood, a routine or an interest, according to YouTube’s Made On announcement and coverage from 9to5Google and Engadget.
The same event added three more pieces that matter to marketers. Ask YouTube, the Gemini-powered search experiment, now builds side-by-side product comparison tables from several videos and web pages, with video titles and channel names shown prominently (Android Headlines). Studio tests are growing from titles and thumbnails to whole video cuts. And dynamic thumbnails will rotate preview images by audience segment (Tubefilter). YouTube hasn’t given a date for the testing tools.
How does a prompt-built feed decide which videos to show?
YouTube hasn’t published how it works. The practical model: the prompt narrows the candidate pool by topic first, and then YouTube’s usual satisfaction signals rank what’s left. That order matters, because a video that never enters the pool never gets a chance to be ranked.
Here’s the part you can predict from. Today your videos get recommended mostly by behavior: people who watched X also watched you. A prompt feed adds a step before that. The system has to turn “furnace troubleshooting before winter” into a set of eligible videos. The only evidence it has for that step is what your video is about: the spoken transcript, title, description, chapters, and whether your channel keeps showing up on the same topic. So a channel with 80 videos where 60 cover furnaces, boilers and heat pumps reads as a clear match. A channel that mixes furnace tips with holiday party clips and hiring posts reads as a weaker one, even if one furnace video has great retention.
Follow that logic and you can predict a case this article doesn’t cover. A Grand Rapids office-furniture maker whose videos all say “workspace solutions” but never actually say “standing desk” out loud will miss a feed prompted with “standing desk reviews.” Its behavior signals might be strong. It just isn’t in the pool.
Does this actually matter for a small business channel?
Yes, but not because feeds will flood you with views on launch day. Most viewers will keep using the default homepage for a long time. The real change is that YouTube is teaching viewers to describe intent, and small specialist channels are good at matching described intent.
YouTube presents Custom Feeds as a way to delight viewers. Keep some skepticism about who benefits. Look at YouTube’s examples: moods, routines, “new creators in a certain niche.” Those are low-commercial-intent prompts. On a feed like that, a general home-improvement creator with 2 million subscribers competes against your 4,000-subscriber channel on production quality, and you lose. The prompts you can win are narrow and local: “Michigan ice fishing gear,” “CNC setup tips for aluminum,” “what to check on a boat before winter storage.” Our read is that the benefit goes to channels that are narrow and consistent, not channels that are big. The losers are generalist brand channels that treat YouTube as a place to park commercials.
Here’s what’s at stake. Say a Dearborn plumbing company spent $6,000 last year on twelve polished videos: six brand spots and six how-tos. Only the six how-tos have a topic a prompt can match. Half the budget is invisible to the new surface. The fix is to reshoot for clarity, not to spend more.
How should a three-person team make its channel prompt-ready?
Audit topic consistency, make the spoken words match how people search, and use Studio testing to learn which version holds attention. Here’s the weekly version:
- Write down the five prompts you want to win. Phrase them the way a customer would type them, not the way your sales deck talks. A Rochester Hills roofing firm might write “ice dam prevention for older Michigan homes.”
- Map every public video to one of those five prompts or to “none.” Anything marked “none” that has no real viewing value (old event recaps, hiring clips) goes to unlisted or into a playlist that sits apart from your core topic playlists.
- Fix the first 30 seconds of speech. Say the topic out loud in plain words: “If your furnace short-cycles in October, here are four causes.” Then check that YouTube’s auto-generated captions spell your terms correctly, and upload corrected captions where they don’t.
- Add chapters with query-shaped labels. “Why the flame sensor fails” beats “Part 2.” Chapters give a prompt-matching system clean segments to point at.
- Organize playlists by prompt, not by date. One playlist per target prompt shows that your channel covers the topic, not that you once touched it.
- Use Test & Compare now, and full-cut tests when they arrive. Title and thumbnail tests already exist. When up to three video cuts become testable, test the opening first. For explainer channels, the intro is where most viewers leave.
- Record a baseline this week. Export last 28 days of traffic sources from YouTube Studio so you can tell whether a new “feed”-type source shows up after launch. Don’t assume YouTube will label it clearly on day one.
If you’re also running Google Ads video, check how automated creative changes interact with this work. We covered the risks in our look at Performance Max AI video reframing.
Why does Ask YouTube’s comparison table matter more for product sellers?
Because it answers a buying question directly, pulls from several videos and web pages, and credits the source channels by name. That’s an answer-engine surface inside YouTube, and it rewards clear, factual product content over hype.
Take a Warren aftermarket parts seller. A comparison table for “best cold-air intake for a 5.7 Hemi” is only useful if the source videos actually state fitment, part numbers and model years. Say them out loud, and put them in the description too. YouTube also said shopping tags drive 2x more clicks than links in descriptions (reported by TheWrap), and that the Shopping affiliate program will reach 35 countries by year-end. If your products are tagged, a table mention has a short path to a sale. The same fitment discipline applies to feeds, as we explained in our Merchant Center product video mapping audit.
What should Michigan brands watch on the calendar?
The next eight weeks line up YouTube’s rollout with Michigan’s busiest search season for cold-weather services.
- September 23, 2026: Made on YouTube announcements. Shorts series (seasons and episodes) starts rolling out.
- October 2026 (reported, unconfirmed by YouTube): Custom Feeds reach US viewers on web, mobile and TV, per Android Headlines. YouTube’s own wording is “soon.”
- Mid-October through November: Michigan furnace tune-up, winterization and boat-storage demand peaks. Have your prompt-mapped playlists live before this.
- Year-end 2026: YouTube Shopping affiliate program expands to 35 countries.
- Early 2027: Live auto-dubbing pilots and co-streaming matchmaking arrive (TheWrap).
Is TV viewing part of this story?
Yes. Custom Feeds is launching on TVs as well as phones, and the living room is where a lot of YouTube viewing now happens. YouTube’s Kurt Wilms told TheWrap that for more than half of the top 100 creators, TV is now the most-watched screen, and viewers watch 2 billion hours of Shorts on TVs each month.
For a Michigan contractor, that changes the creative. A homeowner in Southfield watching on a TV can’t tap a link. Put your company name and service area in the video itself, on screen and said out loud, and don’t count on the description to do that work.
Frequently asked questions
Can I target Custom Feeds with ads?
YouTube has not announced any ad targeting tied to Custom Feeds, and nothing at Made On 2026 described one. Assume for now that feeds are an organic surface. Any claim that you can buy placement in a specific prompt-built feed is unconfirmed. Keep paid YouTube campaigns on your current targeting, and treat feed visibility as a result of channel structure and content clarity.
Should I rename my old videos to match prompts?
Rename selectively. Change titles on evergreen how-to videos where the current title is vague, and test the change with Test & Compare instead of rewriting everything at once. Leave titles alone on videos that already pull steady search traffic. For most of your library, the spoken transcript, captions and chapters carry more topical meaning than the title does.
Will Custom Feeds launch outside the US?
YouTube announced Custom Feeds for US viewers on web, mobile and TV, with no international timeline yet. Michigan businesses that sell into Ontario should assume Canadian viewers won’t see prompt-built feeds at first. Cross-border channels should keep watching standard search and suggested traffic as the main discovery sources for Canadian audiences until YouTube confirms more.
How many videos does a channel need to show up in a feed?
YouTube hasn’t published a minimum. The sensible target is depth on each topic, not total volume: at least eight to ten videos that clearly cover each prompt you want to win, grouped in a playlist. A channel with ten sharp furnace videos gives a prompt-matching system more to go on than one with fifty unrelated uploads.
Comments are closed