--- title: "What TikTok, Instagram and YouTube say about recommendations" description: "Stop guessing about \"the algorithm\" Creators spend a lot of energy on algorithm rumors: the best hashtag count, the magic posting time, whether editing a…" url: https://optimizeall.com/learn/short-form-video-content-creation/how-recommendation-systems-work updated: 2026-10-05 --- Short-Form Video & Content Creation · Platforms, recommendations and accessibility · lesson 9 of 15 · 14 min # What TikTok, Instagram and YouTube say about recommendations ## Stop guessing about "the algorithm" Creators spend a lot of energy on algorithm rumors: the best hashtag count, the magic posting time, whether editing a caption "kills reach". Most of that is noise. The platforms themselves publish explanations of how their recommendation systems work. They do not share every detail, but what they say is consistent and practical: **show people what they are likely to enjoy, based on what they and similar people actually watch and do.** ## TikTok: the For You feed TikTok says For You recommendations are based on a combination of factors: - **User interactions:** videos you like, share, comment on, watch to the end or skip; accounts you follow; content you create. - **Video information:** captions, sounds, hashtags and other details. - **Device and account settings:** language preference, country setting and device type (weaker signals). TikTok says these factors are weighted by how strongly they indicate interest. Finishing a longer video is a strong signal; being in the same country is a weak one. It also says **follower count and previous high-performing videos are not direct factors**, which is why a new account's video can reach many people. TikTok also works to diversify recommendations and may not recommend some content, such as material that violates its guidelines. **What this means for you:** completion and rewatching matter, so tight edits and strong hooks matter. Clear captions, on-screen text and spoken keywords help TikTok understand what your video is about, which also matters for TikTok Search. ## Instagram: Reels, Feed, Explore and Stories Instagram uses different ranking for different parts of the app. For Reels, Instagram's head Adam Mosseri has said the most important signals are **watch time**, **likes per reach** and **sends per reach** (how often people share a Reel by direct message). Sends matter especially for reaching people who do not follow you. Instagram has also said it favors **original content**: in April 2026 it extended its originality rules so accounts that mostly repost content they did not create, or have not meaningfully changed, are not recommended to non-followers. Visible watermarks from other apps are also discouraged. Useful Instagram features: - **Trial Reels** show a Reel to non-followers first, so you can test ideas without affecting what your followers see. - **Collab posts** co-author a post with another account and show it to both audiences. **What this means for you:** make Reels people watch to the end and want to send to a friend. Upload original, clean files. ## YouTube Shorts and long-form YouTube says its recommendation system aims to help each viewer find videos they want to watch and to **maximize long-term satisfaction**. It uses signals including clicks, watch time, shares, likes and dislikes, and **user surveys** that ask viewers to rate videos; YouTube calls highly rated watch time "valued watch time" and uses models to predict satisfaction for everyone. Since March 31, 2025, a Shorts view counts whenever a Short starts to play or replay, while the older, stricter measure is reported as **engaged views** and is still used for YouTube Partner Program eligibility and revenue. **What this means for you:** titles matter because YouTube is also a search engine, and Shorts that lead viewers to your long-form videos help build lasting relationships. ## The common thread | Platform | Strong signals the platform talks about | Your lever | |---|---|---| | TikTok | Completion, rewatches, shares, comments, follows | Hooks, pacing, loops | | Instagram Reels | Watch time, likes per reach, sends per reach, originality | Shareable value, original uploads | | YouTube Shorts | Satisfaction, watch time, engagement, search | Titles, satisfaction, links to long-form | ## Hands-on: the myth-check routine When you hear a new algorithm rumor, run this routine before changing your workflow: ```text 1. SOURCE Did the platform say this (help center, newsroom, official creator account)? 2. SIGNAL Which published signal would it affect (watch time, shares, satisfaction)? 3. TEST Can I test it with Trial Reels, alternate hooks or two similar posts? 4. MEASURE Which metric decides (retention at 3s, sends, completion rate)? 5. DECIDE Adopt only if my own data improves over several posts. ``` ## Worked example (illustrative) A Riyadh food creator hears that "Reels over 90 seconds get no reach". Instead of cutting everything short, she checks Instagram's own statements (watch time is a key signal, and Reels can now be up to three minutes) and tests: two recipe Reels of 45 seconds and two of 2 minutes, same style. The longer ones get fewer views but more sends and saves. She keeps short Reels for discovery and uses longer ones for her most useful recipes. ## Common mistakes - Changing strategy based on rumors without checking official sources. - Chasing likes when shares, sends and watch time matter more. - Reposting others' content and wondering why reach falls. - Judging a Short only by its views, ignoring engaged views and retention. ## Video lecture: What TikTok, Instagram and YouTube say about recommendations Lecture coming soon · 13 chapters · about 9 minutes. Read the full transcript below. 1. How recommendations work 2. The key idea 3. TikTok For You 4. Instagram Reels 5. YouTube 6. The bookshop owner 7. Example 1: London hairdresser 8. Example 2: Riyadh food creator 9. Watch me: myth-check 10. What won't be recommended 11. Signals and levers 12. Recap 13. Try this now ## Lecture transcript ### How recommendations work Post at exactly six p.m. Never edit your caption after posting. Use exactly five hashtags. Videos over ninety seconds get no reach. Have you heard any of these? Algorithm rumors spread faster than any video. And most of them are noise. Here's the thing: the platforms actually publish explanations of how their recommendation systems work. In this lesson, you'll learn what TikTok, Instagram and YouTube officially say, what it means for your content, and a five-step routine for checking any rumor before you change your workflow. ### The key idea Why does this matter? Because every hour you spend chasing a rumor is an hour you're not spending on better hooks, better stories and better editing, which are the things the systems actually respond to. Here's the key idea. Recommendation systems try to show each person what they're likely to enjoy, based on what they and similar people actually watch and do. The platforms don't share every detail. But what they do share is consistent: real viewer behavior drives distribution. So make things people genuinely want to watch, finish, share and come back to. ### TikTok For You Let's start with TikTok. TikTok says For You recommendations are based on a combination of factors. User interactions, like the videos people like, share, comment on, watch to the end or skip, and the accounts they follow. Video information, like captions, sounds and hashtags. And device and account settings, like language preference, country setting and device type, which are weaker signals. TikTok says strong signals, like whether someone finishes a longer video, get more weight than weak ones. And here's the part creators love: TikTok says follower count and previous high-performing videos are not direct factors. That's why a brand-new account can reach many people with one great video. ### Instagram Reels Now Instagram. Instagram uses different ranking for Feed, Stories, Explore and Reels. For Reels, the head of Instagram, Adam Mosseri, has said the most important signals are watch time, likes per reach and sends per reach. Sends, meaning people sharing your Reel by direct message, matter especially for reaching people who don't follow you yet. Instagram also favors original content. In April 2026, it extended its originality rules, so accounts that mostly repost content they didn't create, or haven't meaningfully changed, aren't recommended to non-followers. Useful features: Trial Reels to test with non-followers first, and Collab posts to reach two audiences. ### YouTube And YouTube. YouTube says its recommendation system aims to help each viewer find videos they want to watch and to maximize long-term satisfaction. It looks at clicks, watch time, shares, likes and dislikes, and it runs surveys asking viewers to rate videos. YouTube calls highly rated watch time valued watch time, and uses models to predict satisfaction for everyone, not just the people who answer surveys. For Shorts, remember the counting change: since March 31, 2025, a view counts whenever a Short starts to play or replay, and the older, stricter measure is shown as engaged views, which YouTube still uses for Partner Program eligibility and revenue. ### The bookshop owner Here's an analogy that ties it together. Think of each platform as a busy bookshop owner who recommends books to customers. The owner watches what each customer picks up, how long they read, what they recommend to friends, and whether they come back happy. The owner doesn't care how famous the author is. They care whether readers love the book. TikTok watches for finishing and rewatching. Instagram watches for watching and sending to friends. YouTube watches for satisfaction. Different owners. Same principle. ### Example 1: London hairdresser A simple example. A London hairdresser hears that Reels over ninety seconds get no reach, so she cuts her tutorials to thirty seconds, and they feel rushed. Then she checks what Instagram actually says. Watch time is a key signal, and Reels can now be up to three minutes. Nothing says long Reels are punished. So she lets her step-by-step tutorials run to about a minute and a half, with tight pacing. They get fewer views than her short clips, but many more saves and sends, and more booking inquiries. ### Example 2: Riyadh food creator Now a realistic scenario, illustrative. Lina is a food creator in Riyadh with twenty thousand followers. Her reach drops, and a friend says it's because she edits her captions after posting. Instead of panicking, she looks at her own data. Her last six Reels were reposts of recipe clips from other accounts with her voiceover. Instagram's originality rules explain the drop far better than a caption rumor. She switches back to filming her own recipes, uses Trial Reels to test two new formats, and within a few weeks her reach to non-followers recovers. ### Watch me: myth-check Watch me run the five-step myth-check routine on a rumor: posting at exactly six p.m. doubles your reach. Step one, source. Did the platform say this? I search the help center and official creator accounts. No. Step two, signal. Which published signal would it affect? Timing could change how many of my followers are online when I post, which might affect early interactions. So it's plausible, but not magic. Step three, test. I'll post four similar videos, two at six p.m. and two at nine p.m. Step four, measure. The deciding metric is early retention and shares per thousand views, not raw views. Step five, decide. I only change my schedule if my own data improves across several posts. ### What won't be recommended A quick word on what the systems won't recommend. Each platform says it may limit recommendations for content that breaks its rules or community guidelines, and some sensitive topics may be shown less widely even when they're allowed. On TikTok and YouTube, repeatedly posting content that isn't original, or that's mass-produced with little variation, can hurt distribution and monetization. So if a video suddenly gets far less reach than usual, check two things before blaming the algorithm. First, did anything in the video break a guideline, like a dangerous stunt or unlicensed music? Second, is it genuinely original and complete? Most mysteries have ordinary explanations. ### Signals and levers Let's put the common thread in one table. TikTok talks about completion, rewatches, shares, comments and follows, so your levers are hooks, pacing and loops. Instagram talks about watch time, likes per reach, sends per reach and originality, so your levers are shareable value and original uploads. YouTube talks about satisfaction, watch time and engagement, plus search, so your levers are titles, genuinely useful content and links to long-form. The common mistakes: changing strategy based on rumors, chasing likes when shares and watch time matter more, reposting others' content, and judging a Short only by views. ### Recap Recap. TikTok says For You is shaped by interactions, video information and settings, and follower count isn't a direct factor. Instagram names watch time, likes per reach and sends per reach, and rewards originality. YouTube aims for long-term satisfaction, including survey-based valued watch time. And before believing any rumor, check the source, find the signal, test, measure and decide. ### Try this now Try this now. Pick one algorithm rumor you've heard recently. Run the five-step myth-check routine in your notes app. Then design a two-week test, using Trial Reels, alternate hooks or paired posts, with one deciding metric written down before you start. In the next module, we'll build a content strategy that balances trends with evergreen content. ## Key takeaways - TikTok says For You is shaped by user interactions, video information and device/account settings; follower count is not a direct factor. - Instagram names watch time, likes per reach and sends per reach as key Reels signals and favors original content. - YouTube aims for long-term satisfaction, using watch time, engagement and survey-based valued watch time. - Check rumors against official sources, then test with your own data before changing your workflow. ## Try it Pick one algorithm rumor you have heard, run the five-step myth-check routine, and design a two-week test using Trial Reels, alternate hooks or paired posts. - [Previous: TikTok vs Reels vs Shorts: what changes](https://optimizeall.com/learn/short-form-video-content-creation/tiktok-reels-shorts) - [Next: Trends vs evergreen: a balanced content mix](https://optimizeall.com/learn/short-form-video-content-creation/trends-vs-evergreen) - [All lessons of Short-Form Video & Content Creation](https://optimizeall.com/learn/short-form-video-content-creation)