Pinterest Trends & Content Intel

TL;DR for marketers, e-commerce / DTC, and content strategists: Pulls Pinterest trend and content intelligence from logged-out public surfaces - trending keywords with interest trajectory + seasonality, top pins per topic (saves / engagement), and board / creator audience signals - normalized into one clean schema. The official Pinterest API gates this behind Trial -> Standard app review plus an OAuth screen-recording (trial is capped at 1,000 requests/day), so most teams never get the trends/content endpoints. This actor hands you the same shape of data with no application, no API key, no login. Public data only, PII-minimized, polite rate limits, fail-soft when blocked. The free Apify plan covers exploration runs on your $5 platform credit.

Run it in 30 seconds

# Via the Apify Python SDK
from apify_client import ApifyClient

client = ApifyClient("<YOUR_APIFY_TOKEN>")
run = client.actor("seibs.co/pinterest-trends-intel").call(run_input={
    "mode": "trend_search",
    "keywords": ["home office decor", "fall outfits"],
    "region": "US"
})
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item)

Or via curl:

curl -X POST "https://api.apify.com/v2/acts/seibs.co~pinterest-trends-intel/run-sync-get-dataset-items?token=<YOUR_APIFY_TOKEN>" \
  -H "Content-Type: application/json" \
  -d '{"mode": "topic_content", "keywords": ["air fryer recipes"], "max_pins_per_query": 25}'

Or click "Try for free" on this page if you prefer the no-code UI.

Why this exists (the gate is the moat)

The Pinterest API is free, but useful access is approval-gated: a new app gets Trial access (capped at 1,000 requests/day) only after review, and moving to Standard access requires a second app review that includes a screen-recording of your OAuth flow plus Developer-Guidelines compliance. DTC brands, agencies, and independent trend forecasters routinely can't or won't clear that gauntlet - yet Pinterest is one of the highest purchase-intent platforms on the web, so the trend + content data is exactly what they need. This actor reconstructs that data from the public Pinterest surfaces any logged-out visitor can see, with no application required.

What you get

Each run produces:

• A clean dataset, filterable in the Apify console and downloadable as CSV or JSON
• Views: overview (AI-agent / analyst skim), trends, pins, creators
• An access_notes record up top documenting which surface each field came from, the anti-bot escalation telemetry, and the responsible-use note
• A sample-output preview at ./.actor/sample-output.json

Modes

Mode	Input	What it returns
trend_search (default)	keywords	One trend record per keyword: interest trajectory (weekly series when the public Pinterest Trends surface is reachable, else a transparent search proxy), seasonality (peak months, seasonal vs evergreen), growth windows (WoW / MoM / YoY), related terms, and category - plus the top pin records behind each term.
topic_content	keywords	A topic_rollup per query (pin count, total/avg saves, top categories, top creators, dominant-color palette, top link domains) plus the ranked top pin records, and optional creator_signal records for the top creators.
pin_detail	pin_urls	Deep normalized pin records (title, description, saves/reactions/comments, engagement score, dominant color, board, creator, outbound domain).
profile_analysis	usernames	A creator_signal per profile: follower / following / pin / board counts, monthly views, and top boards.

The normalized schema

Every surface maps onto one vocabulary so you don't write Pinterest-specific parsing:

• trend - keyword, category, trend_type (growing / declining / seasonal / evergreen), trajectory ([{date, value 0-100}]), trajectory_source, interest_latest/peak/mean, growth_wow/mom/yoy, seasonality {is_seasonal, coefficient_of_variation, peak_months}, related_terms.
• pin - pin_id, title, description, save_count, reaction_count, comment_count, engagement_score (save-weighted), dominant_color, image_url, link_domain, board, creator, category.
• creator_signal - username, full_name, follower_count, following_count, pin_count, board_count, monthly_views, top_boards, category.
• topic_rollup - pin_count, total_saves, avg_saves, avg_engagement, top_categories, top_creators, dominant_colors, top_link_domains.

Sources & access notes

All surfaces are logged-out and public - the actor never logs in, holds no account, and supplies no user cookies (the only cookie it ever uses is the anonymous csrftoken Pinterest itself sets for any visitor, echoed back so the public resource API answers an anonymous browser):

Surface	How it's fetched
Search (topic_content, trend_search)	The anonymous BaseSearchResource JSON API, with a server-rendered __PWS_DATA__ HTML-blob fallback.
Pin (pin_detail)	The pin page's embedded __PWS_DATA__ JSON blob.
Profile (profile_analysis)	The profile page's embedded __PWS_DATA__ JSON blob.
Trends (trend_search)	The public trends.pinterest.com tool - best-effort weekly time series.

Honest note on what each surface exposes. The anonymous search grid reliably returns pin id, image, dominant color, outbound link domain, reaction count, creator (username + follower count), board, and category. Pinterest withholds exact save (repin) counts from the logged-out search API, so engagement_score weights reactions when saves are absent; pin_detail (the PinResource endpoint) recovers richer per-pin fields and, like profile_analysis (UserResource), works most reliably on the RESIDENTIAL proxy + browser tier the Apify platform provides.

Honest note on trajectory. The weekly interest time series is the gated API's prize field. When the public Pinterest Trends surface answers, the trend record carries a real trajectory with trajectory_source: "pinterest_trends". When it doesn't (the Trends tool is JS-heavy and may not expose a stable anonymous series), the record falls back to trajectory_source: "search_proxy" - a transparent interest approximation from search depth + top-pin save velocity - and says so in _trajectory_note. The actor never reports a trajectory it didn't actually fetch.

Anti-bot escalation ladder

Pinterest fingerprints the TLS/JA3 of plain HTTP clients and 403s them from datacenter IPs. On a challenge the client escalates automatically:

1. httpx over the DATACENTER proxy (cheapest, first).
2. curl_cffi with real Chrome TLS impersonation over the RESIDENTIAL proxy - defeats JA3/TLS-fingerprint blocks.
3. patchright / playwright stealth browser over the RESIDENTIAL proxy - for true JS/Cloudflare challenges. Set browser_cdp_url (or the BROWSER_CDP_URL env var) to connect a warm anti-detect browser.
4. Fail-soft - if every tier is blocked, the run emits a documented fetch_error note and still finishes SUCCEEDED.

Per-tier counts are reported in the access_notes.anti_bot_escalation block.

Pricing (pay-per-event)

Event	Price	When
pin_record	$0.004	Per normalized pin emitted.
trend_enrichment	$0.008	Per trend record (trajectory + seasonality + related terms + category).
board_audience_signal	$0.010	Per creator/board signal (profile_analysis, or top creators in topic_content).
scheduled_delta_run	$0.030	Once per scheduled monitor-mode run.

A run that returns nothing costs nothing. The actor also ships a 5-layer cost defense (pre-flight input caps, demo-mode soft-fail, a _RunBudget guard that aborts cleanly if compute outruns revenue, in-loop budget checks, and hardcoded PPE prices) so a runaway run can't bill you for compute it can't justify.

Monitor mode

Save this actor as an Apify Schedule (or "Save as task") and each scheduled run emits a monitor_digest of new/rising trend terms and save/engagement deltas since the last run. Point monitor_webhook_url at a Slack-compatible webhook to get the digest pushed. Scheduled runs charge scheduled_delta_run.

MCP / AI-agent use

Use the companion mcp-pinterest-trends-intel actor for a Model Context Protocol server exposing search_trends, get_top_pins, get_pin, and analyze_profile as agent tools (x402 / Skyfire ready). Or wire the dataset directly: the overview view and the datasetItemsMcp output link return clean, token-efficient JSON for Claude, GPT, LangChain, and LlamaIndex tool calls.

Legal & responsible use

Logged-out public Pinterest pages only - no login, no accounts, no paywalls. Creator usernames and aggregate follower counts are public profile facts; the actor minimizes PII and never enriches into private/personal contact data. Favorable precedent (e.g. Meta v. Bright Data, 2024) holds that logged-off scraping of public data is not barred by platform terms, but you remain responsible for complying with Pinterest's terms and applicable law in your jurisdiction. Intended for trend research, content strategy, and competitive analysis - not bulk content republication. The actor keeps polite rate limits and fails soft when a surface is unavailable.