BusinessRnD · draft v2 · built 2026-10-21 06:15 UTC
I1 · Apify actors — deep dive
Test it
Worth a cheap 4-week test. Big niches are taken; aim at mid-size gaps.Realistic: $2.5–5k/mo in 6–9 months · Chance of $10k/mo: low
Cards have Go deeper, Maybe and Kill buttons. Press Send feedback to save. Your marks from all tabs collect at the end of each page.
Bottom line
The big demand is already taken. The top 100 actors hold 59% of all monthly users. Apify's own accounts hold 39.5%. The ten biggest clusters (Instagram, LinkedIn, jobs, TikTok, Google Maps…) each have 10–65 strong actors and prices down to $0.04–0.35 per 1,000 results. Don't enter them head-on.
The gaps are mid-size niches (1,000–8,000 monthly users) with weak leaders. The Google Trends leader succeeds on 56% of runs. The Google Hotels leader succeeds on 70%. TikTok ads has no leader: the top actor holds 9%. The #1 agent result for Google Ads Transparency is rated 1.93/5.
Newcomers still break in. 32 of the 125 actors with 1,000+ monthly users were created in 2026. A Google Ads Transparency actor launched in June now has 1,112 users. An Instagram reel-transcript actor launched in August has 1,639 at $0.015 per reel.
Agents pick by title first, then popularity. In 55 replicated agent searches, every top-5 result had every search word in its title or slug. 37% of top-5 slots went to actors with fewer than 50 users. Coding agents first read Apify's fixed list of 130 actors, 91 of them Apify's own.
Focus beats volume. harvestapi has 16 actors and 68,920 monthly users. ParseForge built 2,000 actors with Claude Code and Codex and has about 4,300 users: 2.5 per actor.
Money (estimate): about $2.5 net payout per monthly user per month on average. $10k/month means ~4,000 users, which only 28 of 7,172 developers reach. A realistic plan is 6–9 actors and 1,000–2,000 users in 6–9 months: about $2.5–5k/month.
Start with ad-library actors (TikTok, Google Ads Transparency, LinkedIn) plus a short-video transcript actor. They match code you already have and fit the Claude billing rule (no model, or open-source compute billed to the customer).
What the Store data says
Built on 2026-10-20 from the public Store API: 79,909 actors (86% of the ~92,750 listed; the missing tail has near-zero use), details for ~1,560 top actors, and 55 replicated agent searches. "Monthly users" means 30-day users per actor, added up, so customers using several actors count more than once. The derived CSVs are saved with the research notes. Open the mockup
xychart-beta
title "Monthly users by cluster, thousands, Oct 2026"
x-axis ["Instagram", "LinkedIn", "Jobs", "Web crawl", "TikTok", "Maps", "Facebook", "China", "YouTube", "Search"]
y-axis "Monthly users (k)" 0 --> 200
bar [197, 112, 87, 79, 68, 63, 59, 56, 53, 33]
Mid-size niches with a gap
Niche
Monthly users
Leader's users
Leader success / rating
The gap
Reel/TikTok transcripts
7,939
2,035
100% / 5.0
Rising (7-day/30-day ratio 0.57 vs 0.47 store-wide); premium prices of $0.015–0.075 per reel are accepted
Google Trends
4,723
2,452
56% / 3.68
Leader fails almost half its runs, yet Apify's coding-agent skill recommends it
Google Hotels/Flights
4,146
521
70% / 5.0
Google Hotels users up sharply (0.82 ratio); leaders unreliable
Google Ads Transparency
2,899
1,112
99.6% / 5.0
#1 agent result is rated 1.93; leader is 4 months old
AI answers (ChatGPT, Perplexity…)
2,608
279
92% / 3.72
Fragmented; Apify charges $0.20 per query
TikTok ads / Creative Center
2,277
195
95% / unrated
No leader (top actor holds 9%); no Apify actor
App store data and keywords
1,770
216
96% / 5.0
Top holds 12%; "App Store keywords" returns only ≤5-user actors
LinkedIn Ad Library
1,331
361
95% / 5.0
Rising; ~175 runs per user (monitoring = repeat revenue)
quadrantChart
title Demand vs incumbent strength
x-axis Weak incumbents --> Strong incumbents
y-axis Small demand --> Big demand
quadrant-1 Big and taken
quadrant-2 Big and open
quadrant-3 Mid size and open
quadrant-4 Small and taken
Instagram: [0.89, 0.95]
LinkedIn: [0.74, 0.87]
Google Maps: [0.70, 0.83]
Jobs: [0.58, 0.77]
Meta Ad Library: [0.43, 0.67]
Reel transcripts: [0.46, 0.45]
Trends: [0.10, 0.35]
Hotels Flights: [0.23, 0.33]
Ads Transparency: [0.26, 0.26]
TikTok ads: [0.31, 0.21]
LinkedIn ads: [0.22, 0.11]
How to read it: demand is on a log scale from 1,000 to 100,000 monthly users. Incumbent strength mixes the number of strong actors with the leader's share when the leader is good. Trends = Google Trends; Ads Transparency = Google's. AI answers and app-store data sit in the same corner as TikTok ads. No "big and open" niche exists. The opening is bottom-left.
Also in the data
Rising but taken: Douyin (17,477 users), Taobao (13,604) and Naver (7,390) grow fastest, but zen-studio leads with 5.0 ratings and China-specific anti-bot skills.
Weak leaders, risky targets: Amazon reviews has no strong actor (likely a login wall) and the Reddit leader succeeds on 88% of runs. Both carry legal or access risk.
No demand: tenders and procurement: 880 actors, none above 27 monthly users. A Prozorro actor would only be a funnel for I2.
Staleness is not an opening: 98% of actors with 100+ users were updated in the last 6 months.
How agents pick actors
Agents reach actors through three doors.
The Apify MCP server. Its search-actors tool is a plain Store search and returns 5 results by default. Its instructions tell agents to "prefer Actors with higher usage, ratings, or popularity" (Apify MCP docs; source).
Apify's skill for coding agents (Claude Code, Codex, Cursor). It reads a fixed list of 130 actors first ("Apify-maintained (always prefer)") and searches only if nothing fits (actor index).
Apify AI in Console. It uses the same ranking (Apify MCP docs).
Store and MCP ranking track Apify's quality score: reliability, popularity, reviews, ease of use, PPE pricing, limited permissions, your track record, and matching wording across title, README and schemas.
flowchart TD
A["Agent gets a data task"] --> B{"Uses Apify skill"}
D["search-actors with 1 to 3 words, platform plus data type"]
B -->|"yes"| C["Checks fixed list of 130 actors, Apify ones first"]
B -->|"no"| D
C -->|"site not listed"| D
C -->|"listed"| J["fetch-actor-details: README and schemas"]
D --> E["Hard filter: every word in title or slug"]
E --> F["Hidden: no KYC, full permissions, rental"]
F --> G["Ranked by quality: success, usage, rating, schemas"]
G --> H["Top 5 cards show users, rating, price, inputs"]
H --> I["Agent prefers higher usage and rating"]
I --> J
J --> K["call-actor with maxItems and a spend cap"]
K --> L["Runs, success and reviews feed the ranking"]
Agent-selection checklist
Name it the way agents search: platform + every data noun ("TikTok Ads Library & Creative Center Scraper"). "Zillow listings" returns 9 actors with ≤4 users while the 1,184-user Zillow actor, whose title lacks "listings", is missing.
Pass KYC, request limited permissions, use pay-per-event only, no Standby. Otherwise default search hides you (about 17% of popular actors are hidden), the MCP server excludes you, and per-request agent payments are off (Apify monetize docs).
≥99% success on the default input in under 5 minutes. Apify runs that test daily and flags failures after 3 days (testing).
Describe every output field in the dataset schema. Only 25–42% of popular actors do. The hosted MCP shows agents the result fields before they call.
Inputs an agent can fill: few required fields, enums, defaults, maxItems, and clear errors that refuse vague runs.
Use the same words in title, description, README and schemas (the score's "congruency" check).
Price at or below the niche median with tier discounts.
Seed usage and reviews early: popularity and ratings drive the top-5 order (correlation −0.72 within results).
Publish 5–10 task pages per actor, each framed as a goal. Google indexes them and agents get a markdown copy (published tasks).
Ask to be added to Apify's agent-skill list for platforms Apify doesn't cover. The process is undocumented.
Concrete actors to build
I1.1
TikTok Ads Library & Creative Center Scraper
What it is: One actor for TikTok's public EU ad library (advertiser, creative, reach, dates, targeting) and Creative Center top ads and trends, in one clean schema.
Who uses it (people or agents): Growth teams, agencies, app marketers; agents running "competitor ads" or "trend research" tasks.
Evidence of demand: 2,277 monthly users across 153 actors. The top actor has only 195 (9%), no Apify-maintained actor competes, and agent search returns only small, mostly unrated actors.
Why we can win: No leader. Your trend radar already pulls Creative Center. We can win on field-level schema and reliability.
Build effort (AI): 2–3 days of Claude Code, then automatic daily checks.
Compute: No model. Scraping on Apify compute, paid through per-event prices.
Pricing & revenue estimate: $0.0015 per ad (niche median $0.00175), tier discounts. Target 300–500 users by month 6 ≈ $750–1,250/month (estimate).
First step: Pull 50 ads with the trend-radar code, compare fields with the top 5 actors' outputs, publish.
I1.2
Reels, TikTok & Shorts Transcript and Hook Extractor
What it is: Short-video URL in; transcript with word timings, the first-3-seconds hook text, cut times, words per minute and on-screen text out. Your agent or the customer's does the interpretation.
Who uses it (people or agents): Creators, UGC and ad teams, content agents that study winning formats.
Evidence of demand: 7,939 monthly users, rising (7-day/30-day 0.57; 30-day/90-day 0.68). steadyfetch reached 1,639 users in two months at $0.015 per reel. afanasenko has 402 users at $0.075 per reel.
Why we can win: Your InstaDopamine caption and word-sync engine and video-DNA tooling cover the structure fields that plain transcript actors lack. Main competitor: Apify's own reel scraper, which includes transcripts.
Build effort (AI): 4–6 days (media fetch per platform, open-source speech-to-text, scene cuts).
Compute: Open-source speech-to-text running inside the actor on Apify compute, billed to the customer through the event price. No Anthropic, no external API.
Pricing & revenue estimate: $0.006 per transcript plus $0.004 for the structure pack. Target 400–800 users ≈ $1,000–2,000/month (estimate).
First step: Measure compute and bandwidth cost on 100 public reels; publish only if cost stays under $0.003 per reel.
I1.3
Google Ads Transparency Center Scraper
What it is: Advertiser and ad search across Google's Ads Transparency Center: formats, regions, first and last seen dates, creative links.
Who uses it (people or agents): Competitive-intel teams, agencies; agents asked "what ads is X running on Google".
Evidence of demand: 2,899 monthly users. A June-2026 newcomer took 1,112. The #1 agent result is rated 1.93 and holds 602 users.
Why we can win: No Apify-maintained actor, a weak top agent result, and ad-intel overlap with I5.
Build effort (AI): 2–3 days.
Compute: No model.
Pricing & revenue estimate: $0.0015 per ad (median). Target 300–600 users ≈ $750–1,500/month (estimate).
First step: Build it with the same ad schema as I1.1 so agents can chain them.
I1.4
LinkedIn Ad Library Scraper
What it is: Ads from LinkedIn's public ad library: advertiser, creative, format, dates, impressions ranges where shown.
Who uses it (people or agents): B2B marketers and agencies; monitoring agents.
Evidence of demand: 1,331 monthly users, rising (30-day/90-day 0.62). The leader has 361 users and 162,846 runs in 30 days, so usage is repeat monitoring.
Why we can win: A small field. Public ad repository data, not profiles, is lower risk (inference).
Build effort (AI): 2 days.
Compute: No model.
Pricing & revenue estimate: $0.0015 per ad. Target 150–300 users ≈ $375–750/month (estimate).
First step: Ship together with I1.3. Later add a "competitor ads in one call" bundle across I1.1, I1.3 and I1.4.
I1.5
Google Trends Batch Scraper
What it is: Many keywords × regions × time ranges per run: interest over time, by region, rising queries. Built to survive rate limits.
Who uses it (people or agents): SEO and content teams, trend agents (Apify's skill sends coding agents to the 56%-success leader).
Evidence of demand: 4,723 monthly users. The leader has 2,452 users at 55.9% success and a 3.68 rating. One good alternative (data_xplorer: 761 users, 99.5%).
Why we can win: Reliability alone beats the leader, and your trend radar is a head start. Risk: Apify fixes its own actor.
Build effort (AI): 2–3 days.
Compute: No model.
Pricing & revenue estimate: $0.002 per keyword series. Target 300–700 users ≈ $750–1,750/month (estimate).
First step: Stress-test 1,000 keywords for blocking before publishing.
I1.6
Google Hotels & Flights Prices Scraper
What it is: Hotel prices by date with OTA rate breakdown, plus flight fares by route and date.
Who uses it (people or agents): Travel startups, revenue managers, trip-planning agents.
Evidence of demand: 4,146 monthly users. Google Hotels users are surging (7-day/30-day 0.82). The two leaders (521 users each) succeed on 70% and 94% of runs.
Why we can win: Reliability on a growing niche with no dominant actor.
Build effort (AI): 3–4 days.
Compute: No model.
Pricing & revenue estimate: $0.003 per hotel, $0.002 per flight option. Target 200–500 users ≈ $500–1,250/month (estimate).
First step: Check how Google responds at volume. Google is in active litigation against a scraper (Apify legal overview).
I1.7
AI Answer Visibility Checker (bring your own key)
What it is: A brand and prompt list goes in. The actor queries ChatGPT, Perplexity and Gemini through their official APIs using the customer's own keys and returns answers, citations, brand mentions and rank.
Who uses it (people or agents): SEO/GEO agencies, brand teams, marketing agents.
Evidence of demand: 2,608 monthly users. The top actor has only 279 (11%). Apify launched five such actors in Jun–Jul 2026 at $0.20 per query.
Why we can win: Much cheaper total cost than $0.20. Risks: Apify is the incumbent, API answers differ from the consumer apps, and agents rarely carry their own keys.
Build effort (AI): 3 days.
Compute: Customer-side (the customer's provider keys). A managed mode where we pay a provider would need your explicit approval.
Pricing & revenue estimate: $0.01 per prompt check plus the customer's own provider cost. Target 150–400 users ≈ $375–1,000/month (estimate).
First step: Build after batch 1 passes its day-30 gate.
I1.8
App Store & Google Play ASO Keywords and Charts
What it is: Keyword rank checks, autosuggest expansion, difficulty and top charts for both stores.
Who uses it (people or agents): Indie developers, ASO agencies, app-research agents.
Evidence of demand: 1,770 monthly users across 599 actors. The top actor holds 12%. "App Store keywords" returns only actors with ≤5 users. Demand is real but small.
Why we can win: Your ASO CLI and 229k-query autosuggest scanner already do this.
Build effort (AI): 2–3 days (port the CLI).
Compute: No model.
Pricing & revenue estimate: $0.002 per rank check, $0.0005 per suggestion. Target 50–200 users ≈ $125–500/month (estimate).
First step: Only if batch 1 works; it is mainly a cheap reuse of your code.
Plan on about half of these working (estimate): 1,000–2,000 users in total. G2/Capterra reviews (1,595 users, #1 agent result rated 2.69) is the reserve pick.
Apify compute comes out of our payout; prices cover it
Speech-to-text in I1.2
Customer-side (platform compute billed to customer)
Customer, via per-event price
Open-source model on Apify; no Anthropic, no outside API
Interpreting transcripts, hooks, ads
Customer-side
Customer's own agent or LLM
We return structured fields only
AI answer checks in I1.7
Customer-side (bring your own key)
Customer pays OpenAI, Perplexity or Google directly
Non-Anthropic providers; we never hold a shared key
Optional managed AI-answers mode
Approved paid provider
Customer, cost passed through
Off unless you approve it
Agent payments (x402, MPP, Skyfire)
No model
The agent's wallet, via Apify
Apify's own x402 endpoint made $591 in 30 days (x402scan), so treat it as a bonus
Next steps
gantt
title I1 plan from 21 Oct 2026
dateFormat YYYY-MM-DD
axisFormat %d %b
section You
KYC billing and payout setup :y1, 2026-10-21, 2d
section Batch 1
TikTok ads actor :b1, 2026-10-21, 4d
Google Ads Transparency and LinkedIn ads :b2, after b1, 5d
Transcript actor cost test and build :b3, 2026-10-24, 7d
Publish task pages and agent checks :b4, after b2, 3d
Day 30 review :milestone, m1, 2026-11-20, 0d
section Batch 2
Google Trends Hotels and AI answers :c1, 2026-11-21, 14d
Day 60 review :milestone, m2, 2026-12-20, 0d
section Batch 3
ASO actor and ads bundle :d1, 2026-12-21, 10d
Day 120 review :milestone, m3, 2027-02-18, 0d
I do:
Build batch 1 (I1.1–I1.4) with limited permissions, PPE-only pricing, full schemas and agent-search titles.
Before publishing, check each title against the agent search for 2–3 target queries.
Publish 5–10 task pages per actor, watch daily QA, fix breakage within 24 hours.
Re-pull the Store every week and report users, runs, success and revenue against the gates.
You do once:
Apify account billing details, KYC (photo of your ID) and payout by wire to the FOP USD account. KYC is also what makes your actors visible to agents.
Approve the per-actor price list and decide whether the I1.7 managed mode is ever allowed (default: no).
You do recurring:
About zero. Invoices auto-approve on the 14th; add Apify income to the quarterly FOP declaration; read the weekly one-screen report and say go or kill.
Go/kill thresholds (heuristics):
Day 14 per actor: ≥99% success on the daily test and in the agent top 5 for its main query. Otherwise rename or fix.
Day 30 (batch 1): ≥10 outside paying users or ≥$50 earned → batch 2. Below that → kill the weakest two and move to the reserve pick.
Day 60: ≥150 monthly users across actors or ≥$200/month → batch 3.
Day 120: ≥500 monthly users or ≥$1,000/month → keep adding 1–2 actors a month. Otherwise put the portfolio in maintenance mode.
Any actor: fewer than 10 monthly users after 45 days, or negative margin two months running → unpublish.
Risks and open questions
Rented distribution: Apify sets fees, ranking and pricing rules, and can launch its own competitor. It did that for AI answers in Jun–Jul 2026.
Blocking and legal exposure: hiQ v. LinkedIn (2022) and Meta v. Bright Data (2024) favour scraping public, logged-out data, while Google v. SerpApi is ongoing (Apify legal overview). Prefer public transparency libraries, avoid personal data and logged-in scraping.
Price undercutting: prices in big clusters fell to $0.04–0.35 per 1,000 results. Compete on reliability and fields, not price.
Compute eating margin: video bandwidth in I1.2 could cost more than expected. The cost test comes first.
AI factories: ParseForge and others clone popular actors within days. A head start lasts weeks, not years.
Data limits: Apify publishes no search-volume data, revenue or agent share of usage. The $2.5-per-user payout figure is a rough average from Apify's $1.7M/month payout claim.
Open questions: Will Apify add outside actors to its coding-agent list? Does the Store search match README text? (Results suggest title and slug matter most.)