How Tidera Works

Every search on Tidera issues a live query to ClinicalTrials.gov — the official U.S. registry maintained by the National Library of Medicine. We use the ClinicalTrials.gov v2 REST API and filter results to recruiting-only status.

Results reflect what is actually recruiting today, not a cached snapshot. We do not maintain our own trial database. If a trial closes enrollment between your searches, the updated status will appear the next time you search.

Trials are ranked before any AI model sees them. A deterministic scoring engine evaluates each trial on structured signals. Nothing random, nothing learned — the same inputs always produce the same ranking.

The engine scores on:

• Age and sex eligibility — verified against the trial’s stated minimum and maximum ages, and sex eligibility field. Clear mismatches (e.g., a 40-year-old in a 60+ trial) reduce the score.
• Site proximity — trials with sites in your area score higher when you prefer local options. Trials that are international-only score lower for patients who want nearby access.
• Condition match — trials whose stated conditions directly match your search rank above broader or adjacent studies.
• Treatment history — when you provide medications, the engine detects drug classes (biologics, JAK inhibitors, chemotherapy, etc.) and matches them against trial language like “biologic-naive required” or “must have failed prior therapy.”
• Trial phase — Phase 3 and Phase 4 trials receive a modest boost. Phase 1-only trials (first-in-human, safety studies) are ranked lower for patients seeking treatment access.

The engine returns the top 25 results. Only the top 15 are passed to the AI explanation layer. Trials that score poorly don’t consume AI resources and don’t appear in your results.

The top-ranked trials are analyzed by Claude (Anthropic), an AI model that reads each trial’s eligibility criteria and generates plain-English output for your specific situation. The AI receives your condition, age, sex, location, and any medications you provided.

For each trial, it produces:

• A plain-English summary of what the trial is studying
• The factors from your profile that align with — or conflict with — the trial’s criteria
• A screening complexity estimate (simple, moderate, or complex)
• Questions to bring to a research coordinator, written for your specific situation
• Medication considerations and potential washout conflicts where relevant

The AI reads the trial’s published text. It does not have access to your medical records or any information beyond what you entered. Explanations are generated in batches of 6 trials at a time, with results cached so repeat searches are faster.

Each trial receives a fit score from 1 to 10. The score reflects how well a trial’s stated eligibility criteria appear to align with the profile you described. A higher score means fewer obvious conflicts — not that you qualify.

The AI is instructed to use conditional language throughout and never assert eligibility. When your clinical profile is incomplete, the scoring becomes more conservative, not more optimistic. Phrases like “you qualify” or “you will be accepted” do not appear in Tidera output.

When you search, we log three things for aggregate analytics:

• The condition searched
• The location provided
• How many results were returned

We do not log your name, email, IP address, medications, age, sex, or disease severity. Your health information exists only within your browser session while generating results. It is not stored, associated with an identifier, or shared with third parties.

Your search is not stored against your identity. We cannot connect a search to a specific person.