Body
This article outlines the steps and methods users can take to help prevent and detect fraudulent/bot responses in their REDCap projects. These methods are merely recommendations and not a required aspect of a project.
In This Article:
Understanding fraud and bot activity
Fraud comes from both software and people, and each needs different defenses. Assume any public incentive link will attract all five types below.
|
Type
|
What it looks like
|
Strongest defenses
|
|
Simple bots
|
Scripts that auto-fill forms; very fast, identical or random answers, hidden fields filled
|
CAPTCHA, honeypot fields, timing checks
|
|
AI-assisted bots and agents
|
LLM-written open-ended answers that sound fluent but generic; can pass basic attention checks
|
Contextual open-ended questions, live verification, consistency checks
|
|
Professional or repeat responders
|
Real people submitting many times with varied names and emails to collect multiple payments
|
Duplicate detection, identity verification, payment controls
|
|
Ineligible people misrepresenting themselves
|
Real people who answer the screener the way they think you want, to qualify
|
Hidden eligibility criteria, cross-checks, verification calls
|
|
Coordinated fraud rings and click farms
|
Bursts of submissions shared through forums or group chats, often from similar locations or overnight hours
|
Private links, timing and burst monitoring, staged recruitment
|
Common Red Flags
Built-in REDCap survey settings
Turn on reCAPTCHA and a response limit before any public link goes live; they take minutes and stop the cheapest attacks. Settings that are not visible in your project may need to be enabled by a REDCap administrator.
|
Setting
|
Where to find it
|
What it does
|
Limits
|
|
Google reCAPTCHA
|
Survey Distribution Tools > Public Survey Link > "Protect the public survey using the Google reCAPTCHA feature"
|
Shows an "I'm not a robot" check before the public survey opens
|
Public links only, not participant-specific links; a device passes at most once per day; sophisticated bots and human fraudsters get through
|
|
Response limit
|
Online Designer > Survey Settings
|
Closes the survey after a set number of responses
|
Raise it in small increments after reviewing each batch; protects the incentive budget from a flood
|
|
Time limit for survey completion
|
Online Designer > Survey Settings
|
Expires the survey link a set number of days, hours, or minutes after invitation
|
Applies to invited links; limits time to research or game answers
|
|
Survey expiration date
|
Online Designer > Survey Settings
|
Closes the survey at a fixed date and time
|
Useful for short recruitment windows and pausing a compromised link
|
|
Participant List and Automated Survey Invitations
|
Survey Distribution Tools > Participant List; Online Designer > Automated Survey Invitations
|
Sends a unique, single-use link to each known person
|
Requires a contact list or a screening step first; the strongest protection against link sharing and repeat submissions
|
|
Survey Login
|
Online Designer > Survey Settings (Survey Login)
|
Requires participants to enter values already stored on their record, such as date of birth or a study code, before the survey opens
|
Works only when you hold that information in advance
|
|
Survey access codes
|
Survey Distribution Tools > Participant List
|
Lets participants enter a code instead of clicking a link
|
Codes can be shared; treat like a private link
|
|
Conditional Auto-Continue
|
Survey Settings of the first survey (or survey queue)
|
Moves people to the next survey only when logic is met, such as passing screening or challenge questions
|
Keeps the main survey and the incentive off the public front door
|
Survey design techniques
- Use the public link as a study landing page
- REDCap only allows the first instrument to be a public survey, so make that instrument a short screener or interest form rather than the paid survey itself.
- Eligible, verified people then receive a participant-specific link to the real survey through an Automated Survey Invitation (ASI) or sent manually.
- Honeypot fields
- Add one or two plausible-looking fields with the @HIDDEN-SURVEY action tag. Humans never see them, but many bots fill every field on the page. Any record with a value in a honeypot could be suspected bots.
- Use realistic variable names and labels, not "honeypot" or "bot_trap".
- Set up an Alert & Notification that emails the team when a honeypot field is filled.
- Optionally use branching logic to end the survey quietly when a honeypot has a value.
- Timing fields
- Add a text field with @NOW (or @NOW-SERVER) and @HIDDEN-SURVEY at the start of each page or section to timestamp progress.
- Use the smart variables [survey-time-started:instrument] and [survey-time-completed:instrument], or a datediff() calculated field, to compute duration.
- Pilot the survey with your team and set a minimum plausible completion time, such as one-third of the median pilot time. Flag anything faster.
Identity and eligibility verification
A short live contact is the single most effective check against human fraud and AI agents, because it is costly to fake at scale. Match the level of verification to the study's risk and incentive size.
|
Method
|
How to run it in REDCap
|
Notes
|
|
Phone or video call
|
Store call outcomes in a staff-only instrument; send the main survey link only after a "verified" field is set
|
Ask participants to confirm screener answers in their own words; mismatches and refusals are strong signals
|
|
Email confirmation
|
Send the next survey via Participant List or an Alert to the email given, so the participant must access that inbox
|
Also catches typos; does not stop people with many email accounts
|
|
Mailing address check
|
Compare the address with the stated ZIP code and county; consider mailing a physical incentive
|
Mailed incentives deter fraud because addresses are harder to multiply
|
|
Duplicate detection
|
Make name, email, and phone required fields; review a report sorted by these fields.
|
Check for near-duplicates too: same phone with different emails, swapped name order, plus-addressing (name+1@gmail.com)
|
|
Records or registry match
|
Confirm eligibility against clinic records or a registry under an approved protocol
|
Strongest for clinical populations
|
Tips for verification calls
-
Use a short script with open questions ("Tell me a little about how you heard about the study") rather than yes/no prompts that echo the screener.
-
Note the time zone and background noise only as context, never as a sole reason to exclude.
-
Be respectful; many flagged people are real, and a courteous process protects your study's reputation.
Data monitoring and review
REDCap tools for monitoring
-
Custom reports sorted by start time, email, phone, and duration, with the flag fields as columns.
-
Project Dashboards and Alerts & Notifications to email the team when a honeypot is filled, a record is flagged, or daily submissions exceed a threshold.
-
Data Quality rules for impossible or contradictory combinations.
-
Logging to see when records were created; bursts of records created seconds apart are a warning sign.
-
Survey metadata in exports: start and completion timestamps for each survey.
-
Disabling surveys: If bots have become an issue in a project, the surveys can be disabled temporarily in the survey settings.
If possible assign one named team member to review incoming data on a fixed schedule, daily during the first week of any recruitment wave.