YouTube Hid the AI-Agents Breakouts: The #1 Breakout Sat at Position 50

We built youtube-scraper to rerank YouTube search results, channels, videos, playlists, and shorts by breakout potential, momentum, engagement, recency, and relevance, classify every channel into one of eight signal profiles, and surface an attention queue with per-record whyNow reasons. To stress-test the 2026 chatter that YouTube's first page tells you who's actually rising in a hot niche, we ran it across 100 search results for "AI agents tutorial" on 14 May 2026. The #1 breakout sat at YouTube position 50, the actor flagged two channels in YouTube's top 10 as actively decelerating, and the gap between what YouTube's relevance ranker shows and what the breakout signal says is the data this post documents.

The problem: Every few weeks a creator-marketing newsletter or a B2B SaaS blog publishes the "fastest-rising AI agents YouTubers in 2026" thinkpiece off three screenshots of YouTube's first page. The actual position-by-position rerank of the same 100 results tells a different story. YouTube's relevance algorithm rewards lifetime channel authority, packed historical signals, and search-query match. It does not decay-weight, and it does not surface acceleration. The result is a first page that mixes a couple of real breakouts with stable incumbents and channels the rerank classifies as actively decelerating. SDR teams, sponsor-scouting agencies, and content strategists reading YouTube's order as a "who matters right now" signal are getting handed a stale picture and calling the wrong creators.

This post is a documentary audit of one snapshot of the YouTube relevance algorithm against a transparent breakout rerank for one of the most-searched developer-tutorial queries of 2026. Every position, channel, score, and signal profile in the post links back to either the actor's run output or YouTube's own search page. The data is real, the rerank gap is real, and the framing newsletters usually slap on top is wrong.

What is breakout potential? A multi-axis rerank score the actor computes per channel and video, combining 30-day-vs-60-day view velocity, cadence acceleration (uploads-per-month delta), engagement-rate trend, and signal-event density into a 0-100 score with an attached signalProfile classification (one of eight bands including emerging, breakout, stable-authority, viral-fragile, high-engagement-low-scale, decelerating, dormant, unclassified). See the actor's page on Apify for the full input schema and scoring formula.

Why it matters: YouTube's native relevance ranker is correlated with established channel authority, packed historical signal blend, and query match. It does not decay-weight and it does not expose velocity. YouTube's own description of its recommendation system frames the platform as optimising for relevance and viewer satisfaction, not for surfacing acceleration. Creator-scouting workflows that take YouTube's order as a "who's rising" signal under-index breakouts and over-index stable incumbents.

Use it when: scouting fastest-rising creators in a hot niche, sizing a vertical's actual top-of-funnel attention shift, finding the channels journalists and trade press are about to write about before they write about them, or building a watchlist that tracks velocity instead of authority.

Key findings

• The #1 breakout video, @DigitalSpaceport's "FINALLY USEFUL Local Ai Agents", sat at YouTube position 50. A 90,400-subscriber cybersecurity-and-homelab channel with signalProfile: emerging, momentumScore: 95, three pages deep in the live YouTube results.
• Of the top 10 videos by breakout potential, only 2 appeared in YouTube's top 10. The other 8 were at positions 50, 75, 92, 30-something, and below. YouTube's first page captured a 20% hit-rate on the actual movers.
• YouTube's top 10 contains as many decelerating channels as breakouts: 2 of each. Positions 6 (@codebasics) and 8 (@SkillLeapAI) were tagged signalProfile: decelerating by the rerank, alongside the 2 breakouts at positions 1 and 7. A click on YouTube's top 10 is a coin-flip between rising and falling.
• Subscriber count does not predict breakout. The three highest-momentum channels in the cohort have 90,400, 17,800, and 66,800 subscribers respectively. The five biggest channels in the cohort (TED at 27.4M, Intellipaat at 13M, others) all carry attentionPriority: low.
• 2 of YouTube's top 10 came back with broken channel handles: both resolved to @https://www.youtube.com/u/undefined. YouTube's first page for one of the most-searched 2026 dev queries contained channels the search index could not name properly.
• 27 of 70 channels (38.6%) classify as emerging. Only 2 classify as stable-authority. AI-agents tutorial YouTube is a new-entrant market, not a consolidated one. The space hasn't picked its incumbents yet.
• 171 records, 199 seconds, $0.14 compute. One run produced both YouTube's native order and the breakout rerank simultaneously, because every record carries a rerankAxes object with all five ordering positions.
• One channel appeared four times in YouTube's top 100 results. @iamAImaster surfaced at multiple positions for variant phrasings. The actor deduplicates on channelId post-rerank, yielding 70 unique channels from 100 search records.

In this article: The leaderboard · Story A: the hidden top 3 · Story B: YouTube's top 10 is mixed · Story C: subs don't predict breakout · Story D: the broken-handle gap · Story E: a new-entrant market · What coverage gets wrong · Methodology · Caveats · Press lift-out · FAQ

The top 15 breakout leaderboard

The full top 25 channels by momentumScore for the "AI agents tutorial" query is downloadable as dataset.csv from the "Produced by" banner above. The CSV deliberately omits the recommendedAction column. That's the actor's per-record decision-layer output and the commercial product. The top 15 leaderboard, ranked by momentumScore:

Captured 14 May 2026 via the youtube-scraper actor against YouTube's organic search endpoint for the query "AI agents tutorial", US/English, residential proxy, dateRange: this_month, rankBy: breakoutPotential. Italic rows are large-platform/entity channels (institutions, vendors, news brands); bold rows are individual creators in the high-attention-priority band. Note the plateau at exactly momentumScore: 81 from rank 5 onward: that's the one-signal-event default. The differentiated portion of the leaderboard is the top 3 (95 / 95 / 93), then a small step down to rank 4, then a wide plateau.

Story A: the 3 hidden breakouts YouTube buried

The top 3 channels by breakout potential, with their position in YouTube's native relevance order shown next to where the rerank placed them:

All three classify as signalProfile: emerging with attentionPriority: high. All three carry at least three active signal events including channel_acceleration or cadence_acceleration (the actor's internal velocity and posting-frequency triggers). All three sat three pages deep in YouTube's live first-page paginated results for the query on 14 May 2026.

The mechanism is straightforward. YouTube's relevance ranker scores each candidate result on a packed feature blend that includes query-term match, lifetime channel authority, packed historical performance, viewer-satisfaction signals, and a small recency component. The rerank's momentumScore adds an explicit 30-day-vs-60-day velocity term that YouTube's order does not contain. A channel that jumped from 12 uploads-per-month to 18 uploads-per-month over 30 days, with rising median views, scores accelerating on the velocity axis and gets pushed up the breakout ranking regardless of its lifetime authority. YouTube's first page does the opposite. It weights lifetime authority and packs the established channels at the top.

The bottom of YouTube's first page (positions 40-50) is where the actual movers were sitting on 14 May 2026. The top of the first page was occupied by something else, which is what Story B walks through.

Story B: 8 of YouTube's top 10 are not breakouts

The position-by-position cross-rank of YouTube's top 10 search results for "AI agents tutorial" on 14 May 2026, with the breakout rank in parens and the actor's signal profile:

Two breakouts (positions 1 and 7), two decelerating channels (positions 6 and 8), four unclassified middle-of-pack (positions 2, 3, 4, 5), and two with broken handle resolution (positions 9, 10). The share of decelerating channels in YouTube's top 10 (20%) exactly equals the share of breakouts (20%). A buyer scouting fastest-rising AI-agents creators by clicking the first few YouTube results is as likely to land on a decelerating creator as on a breakout.

This is the documentary core of the post. YouTube's top 10 is not "the top 10 movers in this niche right now." It is a relevance-ranked mix of incumbents, decelerating channels, two genuine breakouts, and a couple of records the search index can't name properly. The breakout signal has to be computed externally because YouTube doesn't expose it.

The top 10 by breakout potential pulls the actual rising channels from across the 100-result window. Two of those overlap with YouTube's top 10 (positions 1 and 7, breakout-ranked 4 and 5). The other eight sit at YouTube positions 50, 75, 92, and similar. Across the full top 10 by breakout, the average YouTube position is roughly 35, three to four pages deep in YouTube's live results.

Story C: subscriber count does not predict breakout

A creator-scouting workflow that sorts by subscriber count surfaces the wrong creators when the question is "who is accelerating right now." The five biggest channels in the cohort by subscriber count:

None of the mega-subscriber channels in the cohort carry a high attention priority. The three highest-momentum channels carry 90,400, 17,800, and 66,800 subscribers respectively. The breakouts live in the 17k-90k subscriber band on this query.

The intuition that more subscribers means more relevance to a niche is wrong here. TED and OpenAI clear 1M+ subscribers because they are large-platform brands across all topics, not because they are dominant on the AI-agents tutorial query specifically. Their relevance to this exact query is no better than a 90k-subscriber homelab channel actively shipping local-AI-agents content every week.

For an SDR team scouting creators for a sponsored video on an AI-agents tutorial product, the 90k-subscriber channel with attentionPriority: high is the right first call. The 27M-subscriber institutional channel with attentionPriority: low is not. The actor surfaces this distinction directly. The conventional spreadsheet sort by subscriber count does not.

Story D: 2 of YouTube's top 10 had broken channel handles

YouTube positions 9 and 10 ("Full Walkthrough: Workflow for AI Coding (Matt Pocock)" and "Amazon Bedrock for Beginners (From First Prompt to AI Agent)") both came back with channelHandle: "@https://www.youtube.com/u/undefined". The actor's handle resolver fell back to a URL string. The video records still resolve correctly via the video ID, but the channel-level scoring is unavailable for these two records.

This is logged as residual actor feedback (a fix is pending). It's also a quiet but real signal about YouTube's first page: two of the top 10 results for one of the most-searched 2026 dev-tutorial queries returned channel handles the actor's resolver, working from the live YouTube DOM, couldn't parse into a clean handle string. YouTube's own search index is not a clean source of channel identity for every result it surfaces.

For documentary purposes the post names the videos but flags the handles as unresolved. A journalist citing the cohort can lift the two video titles and the count (two of ten with broken handles), but should not cite either as a "rising channel" claim. The channel-level signal is missing for both.

Story E: 27 emerging vs 2 stable-authority

signalProfile distribution across the 70-channel cohort (after dedup by channelId from the 100 search records):

The two stable-authority channels are @futurepedia_io and @GitHub. The actor classifies stable-authority as channels with consistent audience, established presence, no significant acceleration, and no significant deceleration: the "incumbent who's holding ground" band. Two channels meet that bar on this query. Twenty-seven hit emerging, newer channels with at least one active signal event.

AI-agents tutorial YouTube is a new-entrant market. The space hasn't consolidated yet. Twenty-seven emerging creators are competing for a top slot that two established channels currently occupy. This is the shape of a niche that's about to undergo a power-law concentration event over the next 6-12 months as a small subset of those 27 emerging channels accumulates the cadence, velocity, and audience signals to graduate into breakout or stable-authority bands.

Compare with our sister audit yesterday on the developer-creator economy: 90 dev YouTubers, only 2 hit A-tier sponsor-readiness, a 2.2% top-of-pyramid. Both audits return the same shape: a tiny top of the pyramid, a long middle, a long tail. Both also return the same operational implication: the conventional sort (by subscriber count, by search-result position) doesn't surface the small set of creators that actually matter for the question being asked. The actor's classifier does.

What most coverage gets wrong about YouTube rising creators

• "YouTube's first page tells you who's rising in a niche right now." Not for this query, not on this date. YouTube's top 10 captured 2 of the 10 top-breakout videos (a 20% hit-rate) and mixed in two decelerating channels at positions 6 and 8. The first page is a relevance-ranked snapshot, not a velocity snapshot.
• "More subscribers means more relevance to a tight niche." Mega-subscriber channels on this query (TED, Intellipaat, OpenAI, IBM Technology) all carry attentionPriority: low because their cross-topic authority doesn't translate to acceleration on AI-agents tutorial content. The 90k-subscriber homelab channel actively shipping in the niche is the rising signal.
• "YouTube's relevance ranker decay-weights old content." The actor's signal-event detection finds active acceleration on emerging channels that YouTube's order doesn't surface. The relevance ranker does include a recency component, but it doesn't expose a velocity score, and the cohort-scale evidence here is that recency alone doesn't pull the breakouts to the top.
• "Sorting by view count surfaces the rising channels." The single best-performing video by raw view count in the cohort isn't necessarily on the breakout list. View count is a stock; momentum is a flow. The two answer different questions. Stories about the fastest-rising creators need the flow signal, not the stock signal.
• "YouTube search results are clean records." Two of the top 10 returned broken channel handles. That's a small but real reminder that YouTube's first-page result set is not a structured database; it's a rendered page with edge cases.

Methodology

• Tool: youtube-scraper build 1.x (post bug-fix cycle, 14 May 2026), run via the Apify platform. The actor ranks YouTube search results, channels, videos, playlists, and shorts by breakout potential, momentum, engagement, recency, and relevance, classifies channels into eight signal profiles, and emits a rerankAxes object on every record so all five ordering positions are present simultaneously.
• Query: "AI agents tutorial". Picked as the dominant 2026 dev-tutorial query, jointly the fastest-growing dev-tutorial niche on YouTube per the OutlierKit 2026 trending-niches report and the ThoughtLeaders 2026 niche-tracking work.
• Run parameters: maxResults: 100, rankBy: breakoutPotential, uploadOrder: relevance, dateRange: this_month, country: US, language: en, proxy: Apify residential US.
• Run ID: VChdyQ94NkW9Ldglq. Dataset ID: Jkrqp1ZJBXieCmvfd. Runtime: 199.1 seconds. Platform compute cost: $0.1387.
• Records returned: 171. Composition: 100 searchResult records (each carrying the rerankAxes object with byRelevance, byBreakoutPotential, byMomentum, byEngagement, byRecency positions), 70 enriched channel records (each carrying momentumScore, momentumBand, signalProfile, attentionPriority, signalEvents, momentumDrivers), and 1 summary record.
• Aggregation rule: dedup channels by channelId after enrichment, yielding 70 unique channels from 100 search records. A handful of channels (@iamAImaster surfaced 4 times, @DavidOndrej and @JulianGoldieSEO twice each) appeared multiple times in the search results across variant phrasings.
• YouTube's native order: captured as the byRelevance axis on every searchResult record. This is the ordering YouTube returned for the query on the actor's residential-US connection on 14 May 2026.
• Breakout rerank: captured as the byBreakoutPotential axis on every record. The actor computes this from a velocity-weighted signal blend including momentumScore, signal-event density, channel cadence acceleration, and engagement-rate trend.
• Known gaps:
  • The recentVideos[].viewCount and channel.totalViews fields are null on this build (residual emit-path bugs flagged after this run). Momentum scoring is computed from internal data and is unaffected, but readers can't independently verify per-video view math from raw.json.
  • Two of YouTube's top 10 returned broken channel handles (@https://www.youtube.com/u/undefined). Video-level records still resolve via video ID; channel-level scoring is unavailable for those two records.
  • Roughly 30 channels in the cohort cluster at exactly momentumScore: 81, the one-signal-event default plateau. The differentiated portion of the leaderboard is the top 3, then a small drop, then a wide plateau.
  • dateRange: this_month was passed but YouTube's interpretation of "this month" surfaced some videos outside the requested window. The actor passes the parameter to YouTube and trusts the returned set.
  • Search-mode results lean US/English. A second-language tutorial market would surface different breakouts.
• Median-views-delta percentages reported by the actor's signal events (for example "Median views rising 160% over 60d") are internal-computation outputs not independently verifiable from raw.json on this build. They are not quoted as standalone claims in the post body.
• Cross-reference for sanity-check: YouTube's live organic search page for "AI agents tutorial" on 14 May 2026. The breakout-rank #1 video, @DigitalSpaceport's "FINALLY USEFUL Local Ai Agents," is a real video at YouTube position around 50 in the live results. Readers can verify the YouTube-side positions by running the same query and counting position numbers.

Caveats and what this data does not say

• One query, one snapshot, one moment. "AI agents tutorial" on 14 May 2026, US/English, residential proxy. The breakout picture for "Claude Code tutorial", "vibe coding", "AI faceless content", or "n8n AI workflow" would surface a different set of channels. Generalising from this single audit to "YouTube hides breakouts on every query" is an over-reach.
• YouTube's relevance algorithm is not failing at its actual job. YouTube optimises for viewer satisfaction and packed signal blend, not for surfacing acceleration. The first page works well for the question YouTube is answering ("which existing videos best match the query"). It just doesn't answer "who's rising in this niche right now." This audit documents the gap, not a YouTube bug.
• Per-video view counts are not independently verifiable from raw.json on this build. The emit-path bug means the recentVideos[].viewCount and channel.totalViews fields are null. The post does not quote any specific per-video view count for this reason. Momentum-score outputs computed by the actor are reported, but readers can't reverse-engineer them from raw view counts in this snapshot.
• The signalProfile: unclassified band is large (45.7%). That's expected. Channels need at least one active signal event to graduate out of unclassified. For channels that haven't crossed the velocity, cadence, or engagement threshold to trigger an event, unclassified is the correct output. It doesn't mean the channel is dead; it means the rerank doesn't have enough signal to classify it.
• The momentumScore: 81 plateau is the one-event default. Roughly 30 channels cluster at exactly 81 because they triggered exactly one signal event. The differentiated portion of the leaderboard is the top 20 or so where multi-event channels separate from the pack.
• No baseline. This is a one-shot snapshot, not a longitudinal audit. The natural temporal anchor is the actor's watchlist mode: scheduled runs against the same query produce week-over-week deltas in momentumScore, signalProfile, and the top-of-leaderboard composition. A 30-day follow-up against the same query would produce the year-over-year-equivalent data.
• The cohort scope is the top 100 organic search results. Videos that exist on YouTube but didn't surface in the first 100 results for this exact query are excluded. A broader query (or a different query in the same niche) would pick up a different sample.

Redaction policy

The public dataset.csv includes the top 25 channels with handle, subscriberCount, momentumScore, momentumBand, signalProfile, and attentionPriority. It deliberately omits the recommendedAction column, which is the actor's per-record decision-layer output and the commercial product. The raw.json is a verifiable subset of the run, not the full output: the same top 25 channels (with recommendedAction stripped from every record), the top 20 search results by YouTube's native relevance, and the top 10 search results by breakout potential, plus the single summary record. 52 records total from a 171-record run. The cuts give readers enough records to verify every claim documented in the post but do not reproduce the actor's full attention queue or per-record rerank output for this query. Snapshot freshness is the second cannibalisation layer: the watchlist deltas are the recurring commercial value, and a single-day baseline is the snapshot.

Press lift-out for journalists

A 2026 ApifyForge analysis of the top 100 YouTube search results for "AI agents tutorial" on 14 May 2026 found that the #1 video by breakout potential, @DigitalSpaceport's "FINALLY USEFUL Local Ai Agents", published by a 90,400-subscriber homelab channel, sat at YouTube position 50 in the live relevance order. Of the top 10 videos by breakout potential, only 2 appeared in YouTube's top 10 search results. YouTube's top 10 itself contained 2 channels the actor classified as actively decelerating, alongside the 2 breakouts. Across the 70-channel cohort, 27 channels classified as emerging and only 2 classified as stable-authority, indicating a new-entrant market that has not yet consolidated. The full leaderboard and methodology are documented at ApifyForge.

Source: youtube-scraper actor, build 1.x, run 14 May 2026, query "AI agents tutorial". Cross-reference: any individual channel's signal profile, momentum score, and breakout-rank position is reproducible by running the same query against the actor with rankBy: breakoutPotential; YouTube's native position numbers are independently verifiable by searching the same query on YouTube and counting result positions.

This post is part of a series of named-cohort ApifyForge audits. See also /data/dev-youtuber-sponsorship-audit-2026 for the sister audit on the developer-creator sponsorship market, and /data/stack-overflow-question-decline-2020-2026 for the broader context of why developer attention is migrating away from Q&A toward video tutorials in 2026.

Embeddable visuals

Chart 1: slope chart of breakout rank vs YouTube position

Slope chart with two parallel vertical axes. Left axis labelled "Breakout rank (1 = top)" with positions 1-10 marked. Right axis labelled "YouTube position (1 = top of first page)" with positions 1-100 marked. Ten lines connect a breakout-rank position on the left to a YouTube position on the right. The top-3 breakout lines (breakout ranks 1, 2, 3) connect to YouTube positions 50, 75, 92: long diagonal lines from top-left to deep-right. The breakouts that appear in YouTube's top 10 (breakout ranks 4 and 5) connect to YouTube positions 1 and 7, flat lines near the top. Title: "Where YouTube placed the top 10 breakouts for 'AI agents tutorial', 14 May 2026." Source line: "ApifyForge / youtube-scraper actor, query 'AI agents tutorial', residential-US proxy, captured 14 May 2026."

Chart 2: stacked bar of YouTube's top 10 by signal profile

Single stacked horizontal bar representing YouTube's top 10 results for the query, segmented by signalProfile. Segments: emerging (2 segments, positions 1 and 7), decelerating (2 segments, positions 6 and 8), unclassified (4 segments, positions 2-5), broken-handle (2 segments, positions 9 and 10). Annotate each segment with the position number and channel handle. Title: "YouTube's top 10 for 'AI agents tutorial': mixed signal profiles, 14 May 2026." Source line: "ApifyForge / youtube-scraper actor, query 'AI agents tutorial', residential-US proxy, captured 14 May 2026."

Chart 3: scatter of subscribers vs momentum

Scatter plot. X-axis: subscriber count (log scale, from 10k to 30M). Y-axis: momentumScore (0 to 100). Each of the 70 cohort channels plotted as a dot, coloured by signalProfile (orange = emerging, blue = unclassified, red = decelerating, green = stable-authority). Annotate the three highest-momentum dots in the top-left corner (@DigitalSpaceport at 90.4k subs / 95 momentum, @AI.with.Hassan at 17.8k / 95, @EmbarkX at 66.8k / 93) and the mega-subscriber low-momentum dots on the right (@TED, @Intellipaat, @himeeshmadaan). Title: "Subscriber count vs momentum, AI-agents tutorial cohort, n=70 channels, 14 May 2026." Source line: "ApifyForge / youtube-scraper actor, dedup by channelId, captured 14 May 2026."

Frequently asked questions

What is the breakoutPotential rerank for "AI agents tutorial" on YouTube?

breakoutPotential is a multi-axis score the youtube-scraper actor computes per channel and video. It combines 30-day-vs-60-day view velocity, cadence acceleration (uploads-per-month delta), engagement-rate trend, and signal-event density into a 0-100 score with an attached signalProfile classification. On the "AI agents tutorial" query as of 14 May 2026, the top three by breakoutPotential were @DigitalSpaceport (95), @AI.with.Hassan (95), and @EmbarkX (93), all classified emerging with attentionPriority: high, all sitting three pages deep in YouTube's native relevance order.

Why did the #1 breakout video sit at YouTube position 50?

YouTube's relevance algorithm scores results on lifetime channel authority, packed historical performance, query-term match, and viewer-satisfaction signals. It does not include an explicit acceleration term. A 90,400-subscriber channel that recently jumped from 12 to 18 uploads-per-month with rising median views scores accelerating on the rerank's velocity axis, but doesn't accumulate enough lifetime authority to clear YouTube's top of the first page. The rerank surfaces the velocity signal; YouTube's order rewards the authority signal. Different optimisation objectives produce different orderings on the same 100-result set.

Are 2 decelerating channels in YouTube's top 10 a bug in YouTube's algorithm?

No, it's a feature of how relevance ranking works. A channel that built lifetime authority through 5+ years of consistent uploads on the topic can still surface near the top of YouTube's results for the query even after its current monthly velocity has slowed. The rerank's decelerating band captures channels with falling 30-day-vs-60-day view trend or shrinking upload cadence. YouTube's relevance ranker doesn't decay-weight aggressively enough to push those channels off the first page on this query. It's an objective-function mismatch, not a YouTube bug.

Does this audit apply to other YouTube niches or just AI agents?

This is a single-query snapshot on 14 May 2026 for "AI agents tutorial". The pattern, YouTube's order under-indexing acceleration and over-indexing lifetime authority, is a structural property of relevance ranking, so it should generalise to other niches where a wave of emerging channels is competing with stable incumbents. The specific numbers (top breakout at YouTube position 50, 2 of top 10 are breakouts, 27 emerging vs 2 stable-authority) would differ across niches. Repeat audits on other hot 2026 dev queries are queued as follow-ups.

Where can I download the underlying data myself?

The top 25 channels by momentumScore are downloadable as dataset.csv from the "Produced by" banner at the top of this post. A verifiable subset of the run is in raw.json in the same folder: 52 records covering the same top 25 channels, the top 20 results by YouTube relevance, the top 10 results by breakout potential, and the summary record. Both files exclude the recommendedAction column. That's the actor's per-record decision-layer output and the commercial product. To get full results with recommendedAction, signalEvents detail, the full per-record rerankAxes on every search result, and watchlist deltas across recurring runs, run the youtube-scraper actor against your own query.

How would a creator-marketing team use this in practice?

The one-shot snapshot is the baseline. The commercial use case is the actor's watchlist mode: set a weekly schedule against the target query, persist outputs, and the actor surfaces deltas like new channels promoted into emerging, channels with rising momentumScore, channels with new signal events, and channels demoted into decelerating. A team scouting AI-agents tutorial creators for a sponsored video next quarter would watch the emerging band weekly and reach out to channels with stable-or-rising momentum scores over a 3-4 week window before they crowd into YouTube's top 10 and become harder to book.

Is this dataset suitable for journalism?

Yes for the cohort-level findings. The top-3 breakout videos and their YouTube positions, the YouTube-top-10 composition by signal profile, the 2-stable-authority vs 27-emerging distribution, and the subscriber-versus-momentum scatter are all reproducible by running the same query against the actor and verifiable by searching "AI agents tutorial" on YouTube directly. For individual-channel velocity claims (for example, "@DigitalSpaceport's median views rose 160% over 60 days"), the per-video view-count emit bug means raw view math isn't independently verifiable from raw.json on this build, and the post avoids quoting those specific delta percentages standalone. The verifiable headline is the position-rank gap, which any reader can reproduce.

Ryan Clinton publishes Apify actors and MCP servers as ryanclinton and builds developer tools at ApifyForge. The leaderboard above was produced via the youtube-scraper actor across 100 YouTube search results for "AI agents tutorial" on 14 May 2026; the methodology, analysis, and framing are independent of any product positioning.

Last updated: May 2026