YouTube Related Videos Scraper

YouTube Related Videos Scraper is the Actor to get YouTube related videos as structured graph edges — recommendation neighbors you can schedule, join, and analyze. Built by ScrapeMamba, this apify youtube scraper (a practical youtube scraper for adjacency maps) calls ScrapeMamba’s POST /youtube/video_recommendations endpoint so teams that need to scrape YouTube related rails get Dataset rows instead of brittle browser scrapes. Use it when you already have video IDs and want recommendation-graph data — not a video downloader or an essay on the YouTube recommendation algorithm.

Actor URL: https://apify.com/scrapemamba/youtube-related-videos-scraper

Pricing on Apify: $2.30 per 1,000 results (billed for successful Dataset rows).

Feed it IDs from Search, Channel Videos, Video Scraper, or your CRM. Chain carefully: recommendation fan-out grows quickly without deduplication.

---

Overview

YouTube’s related / recommended rail changes over time, varies by context, and is not a keyword SERP. Product and research teams still need that adjacency graph: which creators co-occur, which topics cluster around a hero upload, and how “people also watch” neighborhoods drift.

Searches for the YouTube recommendation algorithm usually want either (a) an ML ranking explanation, or (b) a way to export related/recommendation graph data. This Actor answers (b). It does not reverse-engineer ranking models, personalization weights, or feed experiments. It fetches the public recommendations listing for each seed and writes warehouse-friendly rows.

Behavioral facts: endpoint /youtube/video_recommendations; single-page fetch per seed (continuation short-circuits — no infinite related rail). Cap with maxResults (default "40").

Under the hood:

1. videoIds is validated as a non-empty string list (URLs or bare IDs).
2. IDs are normalized and deduped.
3. Up to 8 workers process seeds concurrently.
4. Each seed calls POST /youtube/video_recommendations with { video_id }.
5. Rows whose videoId equals the seed are filtered out.
6. Success rows are type: "related" with sourceVideoId, page, mapped fields, and raw.
7. Per-seed failures become type: "error" with videoId and error.

Integrator note: related hits are type: "related", not "video". Filters that expect "video" will miss them. Keep sourceVideoId — that pair is your directed edge for recommendation-graph analysis.

---

Recommendation graph data (not algorithm reverse-engineering)

What most teams need is an edge list:

• From: sourceVideoId (seed)
• To: videoId (related / recommended neighbor)
• When: run timestamp / Dataset run ID (recommendations are ephemeral)

That is what this youtube data scraper produces. Build adjacency maps, co-recommendation tables, competitor “people also watch” scans, and scheduled drift diffs — without pretending the Actor explains YouTube’s ML. If your research brief says “understand the recommendation algorithm,” translate that into an engineering brief: export the public related graph for these seeds, store edges with timestamps, and measure membership change. That is the problem this Actor solves.

Treat every Dataset as a snapshot of the public related rail at request time. Logged-in browser results, A/B experiment arms, and session personalization can differ from the API snapshot. For product work, that is usually fine: you care about stable join keys (sourceVideoId, videoId) and repeatable schedules more than pixel-perfect parity with one logged-in homepage.

When you recurse (related-of-related), store a global seen-set. Recommendation neighborhoods overlap heavily inside a niche; without dedupe you pay repeatedly for the same videoId and inflate graphs with duplicate nodes.

---

Who this is for

• Analysts building related-video / recommendation graphs from known seeds
• Competitive researchers scanning associations around a hero upload
• Pipelines expanding Search or Channel Videos hits one hop outward
• Teams already running ScrapeMamba youtube scraping Actors on Apify

Who this is not for

• Anyone seeking a YouTube video downloader — this extracts related listing metadata, not media files
• Keyword discovery without seed IDs (use YouTube Search Scraper)
• Deep multi-page recommendation shelf crawls (single page per seed)
• Readers who want an internal tutorial on how recommendation ranking works

---

What it does

• Accepts watch URLs, Shorts-style links (when normalizable), or bare 11-character IDs
• Deduplicates seeds before fetching
• Returns related / recommended listings per seed
• Caps neighbors with maxResults (default 40)
• Tags every success row with sourceVideoId and type: "related"
• Excludes the seed’s own ID from neighbors
• Parallelizes up to 8 seeds
• Records per-seed errors without aborting the batch

---

Use cases

Topic clustering and adjacency maps — Directed edges sourceVideoId → videoId. Cluster titles and channels; re-run on a schedule to watch drift. Export CSV edge lists into NetworkX, Neo4j, or a simple warehouse SQL graph table.

“People also watch” competitive scans — From a competitor hero upload, pull related IDs. Enrich neighbors with Video Scraper; captions with Subtitles when needed. Rank neighbors by how often they co-appear across your seed set.

Expansion after Search or Channel Videos — Expand strong seeds without inventing keywords. Prefer one hop unless you have hard dedupe and budget caps. A common pattern: take the top 5 Search hits, run Related once, then enrich only the union.

Seed lists for enrichment — Collect neighbor IDs, dedupe globally, then batch Video Scraper / Subtitles. Often cheaper than guessing more search queries when you already know a strong canonical video.

Research sampling — Snapshot fixed seed neighborhoods on a cadence; compare week-over-week membership. These are snapshots, not permanent ground truth — recommendations change, and that change is often the signal.

Why ScrapeMamba on Apify — Recommendation rails are awkward to babysit with DIY browsers. This Actor gives you Apify Input/Dataset/schedules, a shared suite join key (videoId), and a clear schema (type: "related" + sourceVideoId) so youtube scraping pipelines stay composable.

---

What data can it extract?

Field	Description
type	"related" on success; "error" on failure
sourceVideoId	Seed that requested recommendations
page	Page index (single page in practice)
videoId	Related video’s 11-character ID
title / author / channelId	When present on the listing
views / length / publishedTime	When available
description / text	Short text when present
thumbnail / url	Preview and watch URL
raw	Original mapped item payload
error	On type: "error" rows (keyed with seed videoId)

Join/dedupe on videoId; use sourceVideoId for graph edges.

---

Input

Field	Type	Required	Default	Description
videoIds	string[]	Yes	—	Watch URLs or 11-character IDs
maxResults	string / number	No	"40"	Max related videos per seed (clamped 1–500)

Prefill example: https://www.youtube.com/watch?v=dQw4w9WgXcQ with maxResults "40".

{
  "videoIds": [
    "https://www.youtube.com/watch?v=dQw4w9WgXcQ"
  ],
  "maxResults": "40"
}

Tips: URLs and bare IDs mix fine. Duplicate seeds collapse before workers run. maxResults is per seed. Smoke-test one public seed with 10–20.

---

Output

Consume success rows as type: "related" + sourceVideoId + mapped fields + raw. Do not expect type: "video".

Success example

{
  "type": "related",
  "sourceVideoId": "dQw4w9WgXcQ",
  "page": 1,
  "videoId": "relatedId123",
  "title": "Related Example Title",
  "author": "Another Channel",
  "channelId": "UCyyyyyyyyyyyyyyyyyyyyyy",
  "views": 120000,
  "thumbnail": "https://i.ytimg.com/vi/relatedId123/hqdefault.jpg",
  "url": "https://www.youtube.com/watch?v=relatedId123",
  "raw": { }
}

Error example

{
  "type": "error",
  "videoId": "dQw4w9WgXcQ",
  "error": "HTTP 500: Upstream request failed after retries"
}

The seed’s own ID is filtered out of success rows even if the listing includes it. Zero valid IDs after normalize → No valid video IDs found. and exit code 1.

---

How to run on Apify Console

1. Open YouTube Related Videos Scraper.
2. Paste videoIds; set maxResults (default "40").
3. Start; confirm Dataset rows show type: "related" and sourceVideoId.
4. Export or pull via Dataset API; dedupe before recursive fan-out.
5. Save as a task + schedule, or call via Apify API.

Checklist: one public seed → confirm type / neighbors → dedupe before scaling.

---

Apify API example

curl "https://api.apify.com/v2/acts/scrapemamba~youtube-related-videos-scraper/runs" \
  -H "Content-Type: application/json" \
  -d '{
    "videoIds": [
      "https://www.youtube.com/watch?v=dQw4w9WgXcQ"
    ],
    "maxResults": "40"
  }'

$curl "https://api.apify.com/v2/actor-runs/RUN_ID/dataset/items"

Client SDKs (JS/Python) use the same Input.

import { ApifyClient } from 'apify-client';

const client = new ApifyClient();
const run = await client.actor('scrapemamba/youtube-related-videos-scraper').call({
  videoIds: ['https://www.youtube.com/watch?v=dQw4w9WgXcQ'],
  maxResults: '40',
});
const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items.filter((row) => row.type === 'related').length);

---

Pricing

• $2.30 per 1,000 results on this listing
• ScrapeMamba usage is separate from Apify compute

Fan-out warning: Re-seeding every related ID without caps grows cost multiplicatively. Prefer low maxResults, single-hop expansion, and global dedupe.

---

Errors and status handling

Retries on 429 / 500 / 502 / 503 / 504 (3 attempts, 2s / 5s / 10s); 60s timeout.

Situation	Behavior
Empty / invalid videoIds	Exit code 1
No valid IDs after normalize	Exit code 1
Per-seed API failure	type: "error"; other seeds continue
Related item equals seed / lacks videoId	Filtered out

---

Best practices

1. Validate with low maxResults first — confirm schema before raising caps.
2. Dedupe ruthlessly across seeds and recursive hops; keep a seen-videoId set.
3. Avoid blind recursion — one hop is enough for many products; multi-hop needs budgets and stop rules.
4. Treat recommendations as ephemeral; re-fetch when freshness matters.
5. Filter on type: "related", not "video". Downstream Video Scraper filters will miss these rows.
6. Always keep sourceVideoId for edge lists and debugging which seed produced a neighbor.
7. Public seeds only — private or unavailable videos error or yield empty neighborhoods.
8. Chain deliberately: Related → Video Scraper for metadata; Related → Subtitles for text; Search for keyword seeds upstream.
9. Schedule weekly snapshots for drift analysis rather than constant live crawling.

---

Limitations

• Single page of recommendations per seed
• Sets change over time and may differ from a logged-in browser
• Seed ID excluded from outputs by design
• Public data only
• Fan-out can explode Dataset size without controls
• Not a media downloader — metadata / listing fields only

---

Troubleshooting

Symptom	What to try
No valid video IDs	Watch URLs or 11-char IDs — not channel URLs or search phrases
Empty for a browser-working seed	Restricted/removed video or unmappable payload; try another public seed
Expected type: "video"	This Actor emits type: "related"; keep sourceVideoId
Huge Dataset	Too many seeds / high caps / recursive re-seed — lower and dedupe
Overlapping neighbors	Normal — unique-set for catalogs; multi-edges for co-rec frequency
Need keywords / captions	Search Scraper upstream; Subtitles Scraper after dedupe

---

FAQ

Does this explain the YouTube recommendation algorithm?

No. It exports related / recommendation graph data (sourceVideoId → videoId) from the public recommendations surface. It does not document ranking ML or personalization.

What Output type do success rows use?

"related", with sourceVideoId, page, mapped fields, and raw. Not "video".

Is this a YouTube video downloader?

No. Like the rest of the suite, this youtube scraper returns metadata and listing text, not video or audio files.

Endpoint, pagination, and defaults?

POST /youtube/video_recommendations with { "video_id": "<seedId>" }. Single-page only. Default maxResults is "40" (clamped 1–500). videoIds required. Up to 8 workers. Seed itself is never returned as a related row.

Why do recommendations differ from my browser?

Personalization, time, region, and experiments differ. Treat output as an API snapshot for the seed ID.

How do I stop cost explosions?

Low maxResults, limit recursion depth, dedupe before re-seeding.

How do I enrich / start from keywords / get subtitles?

Enrich with YouTube Video Scraper. Keywords first via YouTube Search Scraper. Captions via YouTube Subtitles Scraper on deduped IDs.

How do I scrape YouTube related videos on Apify?

Pass videoIds, start the run, confirm type: "related" + sourceVideoId, then export or pull via Apify API / Client (JS or Python).

How is this different from Search?

Search is keyword-driven discovery (queries → listing hits). Related Videos expands from known IDs into recommendation neighbors. Many pipelines use both: Search for seeds, Related for one-hop graph expansion.

Can I use this as a best youtube scraper step for niche mapping?

As part of the suite, yes — especially when your niche is defined by a few hero videos. Related neighborhoods often surface creators you would not find with a single keyword. Combine with Video Scraper enrichment and optional Subtitles for a fuller map.

---

Pipeline patterns

Search → Related → Enrich: Search → pick top IDs → Related (maxResults modest) → global dedupe → Video Scraper (+ optional Subtitles).

Hero neighborhood: One canonical seed → Related once → score neighbors (views, channel overlap, title keywords) → enrich only the top N.

Scheduled graph drift: Fixed seed task → daily/weekly schedule → store edges (sourceVideoId, videoId, runDate) → diff membership over time to see which neighbors stick.

---

Export and automation

Console Dataset download (JSON / CSV / Excel), Apify Dataset API, tasks + schedules, webhooks for enrichment, warehouse loads keyed by sourceVideoId + videoId + run timestamp.

---

Related ScrapeMamba Actors

Actor	Best for
YouTube Video Scraper	Enrich related IDs
YouTube Search Scraper	Keyword seeds upstream
YouTube Subtitles Scraper	Captions on neighbors

Channel, Channel Videos, Channel Shorts, and Trending help when seeds come from uploads or charts.

---

Compliance note

Use only on public YouTube recommendation listings, per YouTube’s terms, applicable law, and your policies. Do not access private videos. You are responsible for how related-video data is stored and used.