What’s New: MTurk closes to new customers, new 'Active in the last x days' filter, and more
From the new ‘Active in the last x days’ filter to demographic breakdowns of participant groups, this month’s updates take your data collection further. Scroll down to explore what’s new.
Amazon MTurk is no longer accepting new customers from 30th July 2026
First-time Prolific researchers can redeem a £50 coupon through 30 July. Find out more
We've also put together a full breakdown of how Prolific and MTurk compare here. These are some of the key differences.
- Data quality: Prolific has consistently delivered higher data quality compared to MTurk across multiple research studies. See the evidence.
- Lower cost per quality response: A recent preprint found that Prolific's cost per high-quality respondent was $5.68 versus $93.33 for MTurk, once data quality is accounted for. Similar analysis by Douglas et al in 2023 found that Prolific cost roughly half what MTurk cost once quality was factored in.
- Prolific is built for quality research, by researchers: We're continually shipping new features to help you run better studies, from Authenticity Checks (to catch AI-generated responses and bots) to Longitudinal Projects (to run multi-wave studies), alongside a steadily growing participant pool.
Know a researcher this could help? Send them straight to the coupon page before 30 July.
Target your study to only active participants
If your study setup typically uses participant groups or saved filter sets, some of the participants who matched last time may be less active today. Until now, there hasn’t been a way to filter for activity directly.
The new ‘Active in the last X days’ filter does this for you. Narrow down the last time a participant was active in the platform, which includes logging in, viewing their dashboard or taking a study. Add it to your study’s filters, combine it with an existing participant group, or save it as part of an existing filter set so your study reaches people those who are genuinely active on the platform.
This means that you can reuse participant groups or saved filter sets without the need to prune or rebuild manually, as less active participants simply won’t be included.
Explore in platform or via API
Know the demographic breakdowns of your participant groups
More Participant Group Hub uplifts this month! Now, you can analyse any participant group directly in the Overview tab. Pick a demographic filter like age, marital status or more and get the percentage and actual count breakdown of your group. Explore different filters for rich insights to learn even more about your participant group.
It means you can sanity-check a group against what your study needs before you launch, especially important if you use groups across teams or studies. Currently available in the UI, with API support coming soon.
Head to the Participants tab in Prolific to explore these updates →
Configure your task layout in AI Task Builder, with more flexibility to come
Where your stimulus sits, which questions appear alongside it and how the screen is split are critical layout decisions that shape how well participants can complete a task. Instead of a fixed structure, AI Task Builder now lets you define the layout for your task.
Each page is built from rows, and each row can hold up to two columns. Put your stimulus in one column and your questions in the next for a clean side-by-side layout, or stack a dataset field above supporting context for something more guided. Each column can hold:
- A dataset field (image, text, audio, etc.)
- An instruction (radio buttons, free text, text with unit selection, file uploads)
- A content block for supporting copy
Mix and match to build side-by-side comparisons, structured annotation tasks, and more.
Unsure how to get started? Explore the new functionality with these ready-made GitHub repos:
- Prolific Smart Glasses POV Video Collection
- AI-Task-Builder-Image-Collection
- AI-Task-Builder-Free-Text-Collection
This is the first step in a series of flexibility improvements in AI Task Builder, watch this space for more! Available now via API for AI Task Builder Batches (annotation tasks).
More verified Experts available for your next study
Our Network reaches over 330,000 90-day active participants so far this month. Standout growth in our Expert Network - our specialist participants individually verified for their skills or expertise.
1,200 new language specialists every month
From major languages to niche dialects, our verified language experts are CEFR-certified across 4 core skills:
- Over 1,000 specialists in major languages such as Spanish (1520+) German (1340+), with Mandarin (890+) closely behind
- Emerging coverage across Telugu (265+), Tamil (180+) and Gujarati (35+). If specific dialects is a priority for your research, explore the variety in our Network today.
Verified healthcare professionals
Our Network now contains over 700 verified doctors across 35 clinical specialities. We cross-check licenses and credentials against official registries to qualify entry. Top specialties in this group currently are paediatrics, internal medicine, emergency medicine and psychiatry.
This segment is actively growing, with other healthcare professions such as nurses (345+), mental health professionals (65+), pharmacists (45+) and more.
If you need professionals from a specific field, find them in-app under Participants > Add screeners > Domain Experts. To find them under the API, take a look here.
Use our Audience Finder to check if your audience is on Prolific.
Watch: New episodes of The Frontier Series
The Frontier Series continues with two new episodes recorded at ICLR 2026 in Rio de Janeiro.
Our AI research engineers Nora Petrova and John Burden sit down with Henry Conklin (Princeton) and Tom Hosking (Cohere) to unpack their paper "Learning is Forgetting: LLM Training as Lossy Compression".
Episode 4 digs into what forgetting reveals about how models learn, and Episode 5 explores what this means for how we train, evaluate and select models.
Use AND / OR logic in saved filter sets
When you need to target a complex audience, AND / OR logic lets you define multiple conditions.
Now, you can apply the same logic to a Filter Set and within nested groups, for more precise targeting (e.g. Employed = full-time OR part-time AND Age = 18-25). This saves you time from manually applying filters each time, especially useful if you run multiple studies with the same criteria.
Quality of life updates this month
- Fetch cleaner message results via the API. If your integration sends automated messages to participants (like return requests or warnings), you can now exclude those from your results when polling
GET /api/v1/messages/. Use the newexcluded_sender_idsparameter to pass a comma-separated list of user IDs, so you only get back genuine participant replies. Find your sender ID, then get started with filtering your messages. - Search all filters from any tab. Search is now available on all tabs (Premium, Popular, All) and searches the entire catalogue at once, not just the tab you’re viewing.
- Streamlining our terminology. “Screener” and “Prescreener” are now called “Filter” in the UI, to align with the language in the API. This means that across the UI and API:
- Filters: the criteria you use to target participants for a study (e.g. age)
- Screeners: a custom screener study you launch to qualify participants before you invite them to your main study
Meet the team in person
Discover our Research at ICML
A busy month for Prolific's AI research scientists as Nora Petrova and John Burden's paper "Pressure Reveals Character" was accepted to ICML. The paper argues that many evaluations test what models know or claim, rather than how they behave when honesty, safety, deference or non-manipulation becomes costly.
The team are also hosting a workshop at ICML: "Theory of Mind and Persuasion Beyond Conversation'' This study introduces a benchmark for evaluating whether LLM agents can plan actions that induce specific belief states in other agents.
We’re heading to ISPP’s 2026 Annual Meeting
Join Dr Andrew Gordon (Staff Researcher - Behavioural Science, Prolific) on 16 July at 10.30am, where he’ll present Not All Samples Are Created Equal: A Comprehensive Comparison of Online Sample Data Quality. Learn more and browse the full itinerary, click here.
Say hello at the CARMA Summer Symposium
On 22 July Dr Andrew Gordon will speak about data quality alongside panellists from Morning Consult and Longwood University from 10am to midday. Learn more here.
Learn about data quality and AI at King’s
Head to King’s Open Research Summer School on 24 July to hear from Emilio Noriega Farres (Senior Product Manager, Prolific) speak about data quality and AI. Register to join the session, or visit our booth for a demo.
Find us at the 2026 Academy of Management Annual Meeting
Register to hear from Dr Andrew Gordon, on 2nd August where he joins the panel at the 2026 Academy of Management Annual Meeting in Philadelphia to discuss Large Language Model Pollution in Online Behavioural Research.
That’s all for our July update! See you next month.
The Prolific team
