Ref sp web social media

Social and Video Platforms

Purpose

Give honest guidance on social and video platforms as discovery surfaces: what the platforms themselves document, what independent research actually shows, and — the part most sources get wrong — how much of the prevailing "social strategy" advice is unvalidated folklore that should not be shipped as a recommendation. This is the counterpart to .agents/skills/ref-sp-web-seo/SKILL.md for the platforms Google does not own.

Verify before you assert — this rots as fast as SEO

Content below was verified 2026-07-13. Platform algorithms, tools, and public statements change constantly, and almost none of it is versioned the way Google's docs are. Every dated claim has a row in ./references/sources.md with its source, its tier, and what would invalidate it. Re-fetch the platform's own newsroom/help pages before asserting a specific — and read the content, not the status code (these pages are client-rendered; a curl may return an empty shell, so use a real browser via .agents/skills/ref-sp-dev-playwright-cli/SKILL.md).

When to use this skill

• Asked whether a brand "should be on TikTok / YouTube / Instagram" for reach.
• Asked how discovery, the feed, or in-app search works on a platform.
• Asked whether posting outbound links hurts reach ("zero-click content").
• Asked how a younger audience finds products, places, or how-tos.
• Asked how social or video content shows up in Google.
• Sanity-checking a social-strategy claim, an agency deck, or a "platform algorithm 2026" article.

Route elsewhere when: the question is attribution or "which channel converts" (.agents/skills/ref-sp-web-marketing/SKILL.md — it owns Direct, dark social, and self-reported attribution; do not restate it here); visibility in AI answers / chatbots (.agents/skills/ref-sp-web-seo-ai/SKILL.md); or ordinary SEO of the site itself (.agents/skills/ref-sp-web-seo/SKILL.md).

The honest position, stated first

Three things are true at once, and most advice collapses them:

1. Platform search is real, and at least one platform documents it. TikTok states plainly that "search powers discovery on TikTok," and ships Creator Search Insights — a first-party tool showing which topics people search and which are "content gaps" (searched often, few videos). That is tier-1 evidence that in-app search matters and a demand-side tool exists. Treat it as the strongest fact in this skill.
2. The headline behaviour stat is real but far narrower than quoted. The famous "40% of young people use TikTok/Instagram instead of Google" traces to a Google SVP describing US 18–24-year-olds "looking for a place for lunch" — "something like almost 40%." It is a real first-party statement about local/dining discovery in one age band in one country. "Gen Z has abandoned Google" is a documented misrepresentation of it. Quote the real version.
3. Almost every tactic layered on top is folklore. Precise reach-penalty percentages, "semantic chunking" of video for LLMs, "transcribe everything," "creator ads always beat display" — these are agency assertions, often from vendors selling the service. Apply the claim-tiering rule from .agents/skills/ref-sp-web-seo/SKILL.md: documented → independently researched → inferred → vendor assertion. Say which tier you are in, every time.

So the deliverable is: name the audience and intent, check whether platform presence is actually warranted for them, do the small documented things, and refuse to ship the folklore.

Is this audience actually on-platform? (do not assume)

The single most expensive mistake is investing in a platform because a statistic said a cohort uses it, without checking that this brand's audience and intent match. Discipline:

• Anchor to real cohort data, and note its bounds. The Swedish Internet Foundation's 2025 survey (real population survey) found TikTok reaches 78% of high-school students and YouTube dominates every age group — strong for that country and cohort, not a global constant. Ask for the equivalent for the actual target market rather than importing a headline.
• Match the intent, not just the demographic. The 40% stat is about finding lunch. Platform discovery is strong for local, product, food, how-to, and entertainment intents; weak for many B2B, high-consideration, or reference queries. A cohort being "on TikTok" does not mean they search TikTok for your thing.
• When the data does not exist, say so. "We do not know whether this audience uses this platform for this intent" is a finding. The fix is a small test, not a strategy deck.

In-app search and discovery

Where a platform documents its own discovery, use that rather than a blog's reverse-engineering.

• TikTok documents search-driven discovery and provides Creator Search Insights (searched topics, content-gap topics, a search-value signal that feeds Creator Rewards) and brand Search Hubs. This is first-party demand data — the closest thing to keyword research the platform offers.
• YouTube is both the second-largest search destination and a video library; its own creator documentation is the source of truth for titles, descriptions, and chapters. Verify current guidance rather than reciting it.
• Recommendation feeds are not search. "For You" ranking is opaque and undocumented; anyone claiming to know its weights is selling a hypothesis. Optimise for the documented search surface, treat the feed as unpredictable distribution.

Outbound links and "zero-click content"

The claim: platforms suppress posts carrying external links, so publish full value natively instead of driving a click. Handle it with calibrated confidence, because the evidence is genuinely mixed.

• The effect is plausible and partly evidenced, but the magnitude is disputed and no platform confirms it. The largest independent LinkedIn analysis (van der Blom, ~1.3M posts) reports a body link reduces median reach ~18–19%; a separate large analysis (Saywhat, ~400k posts) found link-containing posts performed better, crediting content quality. Two big independent studies, opposite conclusions. No platform documents a link penalty.
• So: design as if native value matters, without quoting a number as fact. Lead with standalone value in the post; put the link where it costs least (bio, comment, second step) if the platform suggests it — but present any specific percentage as "reported by X," tier-2 at best, not a law.
• "Zero-click" is a reach tactic, not a measurement excuse. If you keep users on-platform, you also give up the referrer — which lands the visit in Direct. That is a measurement problem owned by .agents/skills/ref-sp-web-marketing/SKILL.md; hand the attribution half there.

Measuring it: two walled gardens

Social measurement fails in two independent ways, and conflating them produces nonsense.

1. Native platform analytics are non-comparable and self-serving. Each platform defines a "view," "reach," and "engagement" differently, does not let you export a clean cross-platform picture, and has an interest in the numbers looking good. Never rank platforms against each other on their own metrics.
2. Off-platform, attribution is broken. A tap from a bio link arrives without a referrer and lands in Direct. This is the dark-social problem — measured, and owned by .agents/skills/ref-sp-web-marketing/SKILL.md (self-reported attribution is the recovery). Do not restate it; route there.

The one first-party cross-surface measurement that exists is Google's Search Console platform properties: connect an Instagram, TikTok, X, or YouTube account and see clicks, impressions, CTR, and average position for how that content performs in Google Search, Discover, and News. Its limit is the whole point — it measures the Google surface only, not native platform engagement (a TikTok video's TikTok views are invisible to it). Useful, and narrow. See .agents/skills/ref-sp-web-seo/SKILL.md for how that Google surface behaves.

Defaults

• Lead with the tier. State whether each claim is platform-documented, independently researched, inferred, or vendor-asserted — the audience is drowning in confident folklore.
• Prefer a small test to a strategy deck. One platform, one intent, a fixed window, self-reported attribution. Learn whether the audience is there before committing budget.
• Do the documented things first: use the platform's own search-insight tool, publish native value, verify current creator guidance. Skip the reverse-engineered feed hacks.
• Re-verify before asserting. Platform tools and statements change monthly; a claim more than a few months old is suspect.

Gotchas

• Do not import a foreign statistic as a local fact. Swedish survey numbers, US survey numbers, and a Google exec's dining anecdote are not interchangeable, and none is global.
• Do not quote an algorithm-penalty percentage as settled. The best independent studies contradict each other and the platforms stay silent.
• Do not treat "For You" reach as controllable. It is undocumented; optimising for it is optimising against a black box.
• Do not confuse native reach with business outcome. A million views that you cannot attribute and that drove no action is a vanity metric — the sample-size and signal-vs-noise discipline in .agents/skills/ref-sp-web-marketing/SKILL.md applies here too.

Scope

Out of scope — paid social advertising. Ad formats, bidding, Spark Ads, and creator-ad buying are paid media, a separate discipline with its own measurement (the platforms provide conversion tracking for spend they are paid for). This skill is about organic discovery and reach. Say so if a user conflates the two; "boost it" is not the same decision as "should we be here."

References

• ./references/sources.md — provenance by section, with each claim's evidence tier and what would invalidate it. Read it before asserting any dated specific.
• .agents/skills/ref-sp-web-marketing/SKILL.md — attribution, dark social, and whether a difference is signal or noise. It owns measurement; this skill routes there rather than restating it.
• .agents/skills/ref-sp-web-seo/SKILL.md — how content performs on the Google surface, and the claim-tiering rule this skill applies.
• .agents/skills/ref-sp-web-seo-ai/SKILL.md — visibility in AI answers, the neighbouring "new discovery surface" concern.