YouTube today published a German-language FAQ setting out how its recommendation systems choose which videos to surface, and the document spends more space on what does not influence distribution than on what does. Alongside it sit two Help Center pages defining the audience reports and the metric vocabulary that creators are pointed toward.

The community post appeared in the German YouTube Help forum under the heading "FAQ: Wie funktionieren die Empfehlungen?" and is attributed to Alyona, a community manager posting under the TeamYouTube account. It carried an Announcement tag and 19 replies at the time of capture. Alyona is the same community manager who published the October 2025 clarification on Partner Program eligibility metrics and the July 2026 Premium Lite market expansion, a pattern in which detailed policy documentation reaches non-English forums first or simultaneously.

Two Help Center articles carry the same footer year and complete the set. One covers the Audience tab in YouTube Analytics. The other is a longer troubleshooting FAQ on discoverability and performance, structured as two collapsible blocks with roughly two dozen questions between them.

The inputs YouTube names

The recommendations FAQ lists three categories of viewer data feeding personalisation. Past preferences, meaning which videos a user has liked before, drive the selection of similar material. Videos watched in sequence by other users create the bridges through which viewers encounter unfamiliar channels. Usage frequency, meaning how often a person watches a given channel or subject, informs what the system serves next.

Popularity is assessed separately. According to the FAQ, the systems examine whether users click a recommended video, ignore it, or press the "Kein Interesse" control, the German rendering of Not Interested. Watch time enters the ranking through two figures: average view duration and average percentage viewed. Likes, dislikes and post-view survey results are folded into the same assessment.

The performance FAQ frames the same machinery differently, describing a candidate set assembled from what viewers watch, what they do not watch, what they search for, their like and dislike ratings, and Not Interested feedback. On the specific weight of ratings, the document hedges. Likes and dislikes are described as belonging to hundreds of ranking signals, and the answer opens with a qualified affirmative rather than a straight yes.

That number, hundreds of signals, is the only quantification of the ranking system anywhere in the three documents. Nothing in the material assigns relative weights, and no signal is described as decisive.

Three external factors

The recommendations FAQ isolates what it calls external factors, meaning conditions outside a creator's control that constrain reach regardless of a video's own performance.

The first is topic interest, defined as the global number of people who watch videos on a given subject. The document states that some topics are more attractive than others and that interest shifts over time.

The second is competition. YouTube systems rank videos from every channel potentially relevant to a viewer, and the FAQ states plainly that a video with good metrics can still receive fewer impressions when other channels post better results. This is the clearest statement in the set that distribution is a zero-sum allocation rather than a reward calculated per video.

The third is seasonality. Traffic fluctuates across the year, and the document points to major public holidays in the countries where an audience lives as a predictable source of variation.

None of the three is quantified. The FAQ gives no indication of how much impression volume a competitive shift can remove, or over what period seasonal swings typically resolve.

What the documents say carries no weight

The larger share of the performance FAQ is devoted to negative findings, and several of them contradict widely held assumptions in creator and agency circles.

Monetisation status

The document states that the search and recommendation system does not know which videos generate revenue and which do not. Videos are recommended on likelihood of viewer satisfaction, independent of monetisation. The FAQ adds one qualification: a video containing graphic or violent content may have monetisation disabled and may also reach fewer viewers, but the reduced distribution follows from the content, not from the monetisation decision.

That distinction matters for anyone reading channel-level data. A correlation between limited ad status and weak reach would, on this account, be a shared consequence of the same underlying content assessment rather than a causal chain running from one to the other.

Channel location

On whether setting a channel's country to a target market improves reach there, the answer is unambiguous. "Nein. Die Standorteinstellungen haben keinen Einfluss auf die Empfehlung von Videos auf YouTube." Location settings have no influence on video recommendation.

Upload cadence and breaks

The FAQ states that the company examined thousands of channels that stopped uploading for a period and found no relationship between the length of the break and changes in view counts. It adds that viewers may take time to resume normal discovery patterns afterward, which is a statement about recovery lag rather than about penalty.

On frequency, the document reports that analyses conducted over years found no connection between increases in view counts and the interval between uploads. Daily or weekly publication is described as unnecessary.

Publishing time

Publication timing is said to carry no influence on a video's long-term performance, with an explicit exception for live streams and premieres, where timing is described as very important. Publishing while an audience is active may benefit initial viewership without affecting the longer trajectory.

Unlisted uploads and title edits

Uploading a video as unlisted and later making it public does not damage performance, according to the document; outcomes depend on viewer response after publication. Changing a title or thumbnail may cause the systems to reassess a video, but the FAQ attributes that to altered viewer interaction rather than to the edit itself. That framing is consistent with the mechanics behind the title and thumbnail A/B testing tools YouTube extended to titles in December 2025, and with the thumbnail governance changes that followed when the company opened Shorts thumbnails while excluding them from testing in July 2026.

Search ranking, described in two lines

YouTube search receives a shorter treatment than the recommendation surfaces. Ranking rests on how closely a video's title, description and content match the query, and on which videos generate the most interaction for that query. A note states that results are not a ranked list of the most-watched videos for a search term.

The document also carries a list of practices to avoid in titles and thumbnails: deceptive or sensational phrasing, shocking or offensive language, repellent imagery, gratuitous violence, suggestive material, and typographic emphasis through capital letters or repeated exclamation marks. Violations may lead to removal under Community Guidelines.

That section is the only place in the three documents where distribution and policy enforcement are described as touching the same surface.

The audience reports underneath

The Analytics page defines the reporting layer creators are directed toward. Monthly viewers counts unique people who watched channel content in the preceding 28 days, recalculated daily on a rolling basis. The page states that the 28-day window persists even when a custom date range is selected in the picker, which means the headline audience figure and the surrounding chart do not always describe the same period.

Viewers are split into three groups. New viewers watched for the first time in the selected period, a category that also absorbs anyone viewing in incognito mode, anyone who deleted their watch history, and anyone returning after more than a year away. Casual viewers watched at least once a month for one to five months in the past year. Regular viewerscleared the same monthly bar for more than six months. Those definitions arrived when YouTube replaced the new-versus-returning split in July 2025.

Nine reports sit in the Audience tab. When viewers are on YouTube covers the last 28 days and is the report the performance FAQ points to for scheduling premieres and live streams. Device type breaks watch time across smartphones, computers, televisions and tablets, with a further split by content type. Bell notifications reports the share of subscribers receiving alerts, a figure whose meaning shifted after YouTube stopped sending push notifications to disengaged subscribers in April 2026. Subscriber watch time splits hours between subscribers and non-subscribers. Age and gender covers demographics.

Three reports describe viewing away from the channel itself. The report on channels an audience also watches uses a 28-day window and is framed as a collaboration prospecting tool. The report on what an audience watches elsewhere uses a seven-day window, covering videos, Shorts, live streams and podcasts consumed off-channel, with a carve-out excluding an official artist channel's own material. The report on formats an audience watches uses 28 days and distinguishes videos, Shorts and live streams. Two further reports cover top countries by watch time and top subtitle languages.

Metric definitions, and one caveat

The same page carries a metrics table. Unique Reach estimates individual people who saw a video or channel and accounts for shared viewing on YouTube TV and the television app, a metric that arrived in Advanced Analytics in May 2026. Views are counted from the start of playback in any format, covering Shorts, on-demand long-form and live streams, which reflects the counting change that took effect on August 24, 2026 and pushed public counters upward without touching the engaged view figures that govern payment. A co-viewing variant of views sits alongside it. Returning viewers and follow-on views complete the list, the latter defined as organic views earned after a person saw promotional content.

One note on the page limits everything above it: "Einige Daten wie Standort, Zugriffsquellen oder Geschlecht können in YouTube Analytics eingeschränkt sein." Some data, including location, traffic sources and gender, may be restricted. No threshold, no explanation of when restriction applies.

Both Help Center pages also carry a disclaimer stating that content may have been machine-translated and that such translations can contain errors. The disclaimer appears on documents that define the metrics creators use to price sponsorships.

Where the material contradicts itself

Two statements in the set sit awkwardly together. The recommendations FAQ opens its section on distribution frequency by stating that content type and format play no role, then immediately describes Shorts feed ordering as a function of popularity and viewer interest, which answers a narrower question than the one posed. The performance FAQ, meanwhile, states repeatedly that not all content is eligible for recommendation on the homepage or in the suggested-video rail, and that videos labelled as made for kids are more likely to be recommended beside other children's content. Format and label plainly do affect placement. The eligibility carve-outs are asserted without specification.

The gap between what the FAQ answers and what creators ask is visible in the thread itself. Several of the 19 replies are channel links attached to requests for help with view counts, one reply asks how to solve the problem, and the remainder are expressions of thanks. None of the questions raised concerns the material the FAQ covers.

Why this matters for media buyers

The documents are a consolidation exercise rather than a policy change, which is precisely what makes their timing readable. They arrive after a sustained run of distribution disputes the FAQ does not address.

Creators documented synchronised viewership declines from August 13, 2025, with desktop traffic falling as mobile rose, and YouTube issued a clarification the following month. A second wave hit Shorts channels from September 7, 2025. In December 2025 the home feed was documented as cutting long-form recommendation slots, reducing them from six per row to two. In January 2026 sort-by-upload-date was removed from search, closing the last route around the ranking layer. The trending page had closed in July 2025.

Set against that record, a document explaining that competition and seasonality move impression volume, and that upload cadence does not, describes a system in which the platform holds every lever the creator does not. That is not a criticism of the explanation. It is what the explanation says.

Three consequences follow for buyers.

The first concerns creator selection. If regular viewer share is the durable measure of an engaged audience, and if the FAQ is correct that publishing rhythm carries no ranking weight, then upload consistency loses its status as a proxy for channel health in partnership screening. The audience segmentation data supports a direct read that the cadence heuristic does not.

The second concerns forecasting. Reach guarantees written against a channel's recent averages inherit exposure to two factors YouTube names and does not quantify. A competitor's strong quarter can compress a partner channel's impressions without any change on the partner's side, and the seasonality warning applies to any campaign crossing a major holiday in the audience's home market.

The third concerns measurement hygiene. A monthly audience figure fixed to a rolling 28 days regardless of the selected date range will not reconcile with a campaign flight of any other length. Combined with the acknowledgement that location, traffic source and gender data may be restricted, and with the machine-translation disclaimer on the page carrying the definitions, the reporting layer supporting creator deals is less determinate than the numbers in the dashboard suggest.

For advertisers buying YouTube inventory directly rather than through creator partnerships, the practical significance is narrower but real. The statement that monetisation status does not influence discoverability establishes that limited-ads inventory is not systematically starved of distribution, which bears on the composition of the addressable pool. The documentation on Kids content filtering published in April 2026 followed the same pattern of explaining an existing system rather than announcing a new one.

What the documents do not contain is any figure describing how much reach any named factor moves, or any account of the platform-side changes creators have spent a year documenting. The word competition covers the first gap. Nothing covers the second.

Timeline

Summary

Who: YouTube, through TeamYouTube community manager Alyona, addressing creators in the German-language Help community, with two accompanying Help Center articles covering YouTube Analytics. The material bears on creators, agencies screening channel partners, and advertisers assessing the composition of YouTube's addressable inventory.

What: A recommendations FAQ naming past preferences, sequential viewing and usage frequency as personalisation inputs, and topic interest, competition and seasonality as the external factors constraining reach. A companion performance FAQ states that monetisation status, channel location settings, upload cadence, upload breaks and publishing time carry no ranking weight, citing an internal study of thousands of channels that found no relationship between break length and view changes. A third document defines the nine Audience tab reports and the metric set behind them, including the rolling 28-day monthly viewers window and the new, casual and regular viewer segments.

When: Published today, September 2, 2026, roughly five hours before capture, with 19 replies on the community thread. The supporting Help Center pages carry a 2026 copyright footer.

Where: The German-language YouTube Help community forum and the German YouTube Help Center. The documented behaviour applies globally.

Why: The material consolidates existing guidance rather than changing policy, arriving after a year in which creators documented undisclosed distribution shifts, the trending page closed, home feed slots moved toward short-form content, and chronological search sorting was withdrawn. It quantifies none of the factors it names, offers no account of platform-side changes, and attaches a machine-translation disclaimer to the pages defining the metrics used to price creator deals.