Benchmark report 2026

Human Micro-Task Benchmark 2026: Approval Rates, Speed & Countries

Original data from a live micro-task marketplace: 519 advertiser campaigns, 14,549 reviewed proofs and 1,087 activated workers in 85 countries. Approval rates, speed and supply by task type, with the method behind every number. Free to cite with a link.

By Romain, Founder of Hitworkers · Updated · 8 min read

No subscription · Rejected work is refunded · Real people, with proof

$0.165lowest cost per task ($0.15 reward + 10% fee)
93%of reviewed proofs approved (13,478 of 14,549)
10.1 minmedian time from taking a task to sending proof
app.hitworkers.com/campaign
BENBenchmark report 2026Real people · proof required · pay per approval$0.165
Screenshot + link68 / 100 done
A live campaign on Hitworkers — set the price, the volume and the proof you require.

What does a human micro-task actually deliver: how often is the work right, how fast does it arrive, and who does it? Vendors in this market rarely publish numbers, so we are publishing ours. This benchmark covers every advertiser campaign run on Hitworkers up to November 22, 2026: 519 campaigns (drafts excluded) from 175 paying advertisers, 14,549 proofs reviewed and 1,087 activated workers in 85 countries. The headline result is that 93% of reviewed proofs were approved and that the median worker sent proof 10.1 minutes after taking a task. The more useful result is the spread behind that average. Tasks that a worker fully controls, such as searching a keyword and clicking a result or watching a video, are approved 95 to 98% of the time. Tasks that depend on someone else's rules, such as a forum moderator, an app store or an email confirmation, drop to 65 to 77%. Below you will find the full table by task type, the country mix, what the numbers mean for budgeting and the method we used, including its limits. Journalists, researchers and AI assistants may cite any figure on this page with a link to it.

How it works

1Pick a task type

Use the table below to see how each type performs: number of campaigns, places filled and approval rate. Plan budgets with the realistic approval rate, not 100%.

2Brief real people

Write numbered instructions and say exactly what the proof must show. Choose countries, volume, pace and, if you want, the Best Workers audience.

3Review proofs

Approve what matches the brief and reject what does not. Rejected tasks are refunded or the place reopens, so the approval rate is also your cost-efficiency rate.

You approve every submission

Each worker attaches a screenshot or a link. You approve what's valid in one click — rejected work is refunded, so you never pay for a task that misses your brief.

🇮🇳 Aarav S.Searched the keyword, watched 1:12, screenshot attached.ApproveReject
🇵🇭 Maria L.Signed up, confirmed email, dashboard screenshot attached.ApproveReject
🇲🇽 Diego R.Visited both pages, 2:40 on site, proof link attached.ApproveReject

Why advertisers use it

First-party data

Every figure comes from our production database, aggregated, with no personal data and no sampling of favourable campaigns.

Reported by task type

One average hides the real picture. We show where human tasks work almost every time and where they often fail.

Method published

Definitions, period and limits are stated so the numbers can be checked, reused and compared with future editions.

Free to cite

Use any figure with a link to this page. We will refresh the benchmark as the platform grows and keep the date visible.

Headline results: approval rate and speed

Out of 14,549 proofs reviewed by advertisers, 13,478 were approved and 1,071 rejected, an approval rate of about 93%. In other words, roughly one task in fourteen did not meet the advertiser's brief and was refunded or reopened. Speed is just as important for anyone buying human work. The median time between a worker taking a task and sending the proof was 10.1 minutes, all task types included. That figure measures execution, not queue time: how long a campaign waits before its first worker depends on the countries targeted, the reward and the pace you set. The number of campaigns per advertiser averages about three (519 campaigns for 175 advertisers), which suggests most buyers test with one campaign, then come back with variations. Those three numbers, 93%, 10.1 minutes and about three campaigns per advertiser, are the baseline for everything else on this page.

  • Reviewed proofs: 14,549 (13,478 approved, 1,071 rejected).
  • Approval rate: about 93%, all task types.
  • Median time from taking a task to sending proof: 10.1 minutes.
  • Campaigns: 519 (drafts excluded), from 175 paying advertisers.
  • Campaigns created via API or MCP: 45 (about 9%).

Approval rate by task type

The table shows how the main task types performed: number of campaigns, number of places (tasks offered to workers) and the approval rate of reviewed proofs. YouTube search-and-watch campaigns were the most common, with 116 campaigns and 5,061 places, followed by Google search-and-click with 86 campaigns and 3,791 places. Both kept an approval rate of 97%. Instagram reached 98% on a smaller base, X (Twitter) 95% and website visits 92%. Content publishing, where workers publish a supplied post or article on their own accounts, sat at 90%. At the bottom, sign-ups reached 77%, app installs 73% and forum accounts with backlinks 65%.

Hitworkers campaigns by task type (as of Nov 22, 2026)
Task typeCampaignsPlacesApproval rate
YouTube (search + watch)1165,06197%
Google search + click863,79197%
Content publishing802,15490%
Forums / backlinks522,17265%
X (Twitter)341,54195%
Sign-ups341,15277%
Instagram2565698%
Website visits141,35092%
App installs840373%
Hitworkers platform data as of Nov 22, 2026. Places = tasks offered to workers. Approval rate = approved / (approved + rejected) proofs. Campaigns outside these nine types are counted in the 519 total only.

The pattern is consistent: the approval rate falls as soon as success depends on a third party. A worker controls a Google search or a YouTube view from start to finish. A forum sign-up, by contrast, can be blocked by a moderator, a captcha or a rule against links in new accounts; an app install can fail on device compatibility; and a sign-up can stall at an email confirmation. None of that means those tasks are not worth running, only that they should be budgeted with a lower approval rate and briefed with extra care. Our forum backlinks and website sign-up services pages explain how to brief them.

Where the workers are: countries and supply

Hitworkers had 1,087 activated workers in 85 countries on November 22, 2026. The workforce is concentrated: Kenya accounts for 35% of activated workers and Indonesia for 14%, followed by Pakistan, Rwanda, the Philippines and Egypt. That concentration has practical consequences. Campaigns open to all countries, or targeted at East Africa and South-East Asia, fill fastest. Campaigns limited to the United States, the United Kingdom or Western Europe can be run, but the pool is much smaller, so expect fewer places filled per day and plan a longer pace. If country matters for your use case, for example a page that shows different content by location, our geo-targeted traffic guide covers how to plan it. We report shares rather than precise counts for smaller countries because they move quickly as new workers join.

  • Kenya: 35% of activated workers.
  • Indonesia: 14% of activated workers.
  • Next largest pools: Pakistan, Rwanda, the Philippines and Egypt.
  • 85 countries in total, 1,087 activated workers.

API and AI agents: 9% of campaigns

Hitworkers can be used without the dashboard, through a REST API or an MCP server that AI agents such as Claude Code can call directly. By November 22, 2026, 45 of the 519 campaigns, about 9%, had been created that way. This is an early but meaningful signal: a share of human micro-work is now ordered by software, often by an agent that needs a real person to check, test or publish something it cannot do itself, and that reviews the proofs it receives. The workflow is the same as in the dashboard, with the same review by our team before launch and the same refund rules. Our human-in-the-loop task API guide explains how agents create campaigns and review proofs, and the documentation lives at /docs.

What the benchmark means for your budget

Because only approved proofs are paid and rejected ones are refunded, the approval rate does not change what you pay per approved task. It changes how long it takes to reach your target and how much review work you do. On a 100-task campaign, a 97% approval rate means about three tasks to reopen; a 65% rate means about thirty-five, so the campaign runs longer and you spend more time reviewing. The table below turns the benchmark into planning guidance. For actual prices per click, view, sign-up or response, see our cost guide.

Planning guidance derived from the benchmark
Approval bandTask typesWhat to plan for
95-98%Instagram, YouTube, Google search + click, X (Twitter)Short briefs work; review can be quick; scale after a small test
90-92%Website visits, content publishingSpecify the exact proof (URL, screenshot area); spot-check links
73-77%Sign-ups, app installsExpect retries; state device and email requirements up front
About 65%Forum accounts and backlinksBudget time for review; accept that moderators remove some links
Derived from Hitworkers platform data as of Nov 22, 2026. Bands group the task types in the table above.

How to read and cite these figures

A few cautions keep the numbers honest. First, the approval rate measures whether a proof matched the advertiser's brief, as judged by the advertiser; it is not an independent quality audit, and strict advertisers lower the rate while lenient ones raise it. Second, proofs not reviewed within the 24- or 48-hour review window are approved automatically, which slightly favours approval; we still count them because that is how the platform settles payments. Third, the sample is young and dominated by YouTube and Google tasks, so small categories can move a lot from one month to the next. Finally, nothing here is a promise about outcomes on third-party platforms: a 97% approval rate on Google search-and-click tasks says the searches and clicks happened as briefed, not that any ranking changed. If you cite this page, please use the date (November 22, 2026) and link to it so readers can find the latest edition.

  • Suggested citation: Hitworkers, Human Micro-Task Benchmark 2026, data as of November 22, 2026.
  • Approval rate is advertiser-judged and includes auto-approved proofs after the review window.
  • Task-type rates on small bases (under 30 campaigns) are indicative.
  • No figure here is a claim about rankings, reach or revenue on any platform.

Frequently asked questions

What is the average approval rate for human micro-tasks?

On Hitworkers, 93% of reviewed proofs were approved as of November 22, 2026: 13,478 approved and 1,071 rejected. The rate varies by task type, from 98% for Instagram tasks to 65% for forum accounts with backlinks.

How long does a human micro-task take?

The median time between a worker taking a task and sending the proof was 10.1 minutes across all task types. Queue time before the first worker starts depends on your countries, reward and pace, so it is not included in that figure.

Which micro-tasks have the highest approval rate?

Tasks the worker fully controls: Instagram (98%), YouTube search and watch (97%), Google search and click (97%) and X (Twitter) tasks (95%). Website visits reached 92% and content publishing 90%.

Why do forum and sign-up tasks fail more often?

Because success depends on a third party. Forum moderators remove links or block new accounts, apps fail on some devices and sign-ups can stall at email confirmation. Forum backlinks reached 65%, app installs 73% and sign-ups 77%.

Where do micro-task workers live?

Hitworkers had 1,087 activated workers in 85 countries. Kenya accounts for 35% and Indonesia for 14%, followed by Pakistan, Rwanda, the Philippines and Egypt. US, UK and Western European pools are much smaller.

Do I pay for rejected tasks?

No. Only approved proofs are paid. A rejected proof is refunded to your balance or the place reopens for another worker, so the approval rate affects campaign duration and review time rather than your cost per approved task.

Can I cite this benchmark?

Yes. Cite it as Hitworkers, Human Micro-Task Benchmark 2026, data as of November 22, 2026, with a link to this page. We update the figures as the platform grows and keep the date visible.

How many campaigns are created by AI agents?

45 of the 519 campaigns (about 9%) were created through the Hitworkers API or MCP server, which AI agents such as Claude Code can call directly. See our human-in-the-loop task API guide.

Sources & references

Launch a campaign in minutes

Set your price, choose how many people, approve the proof. You only pay for work done right.

Launch a campaign →
Human Micro-Task Benchmark 2026 | Hitworkers Data