Skip to content
TruthVouch Docs

Scheduling Cross-Checks

Schedules control when and how often Shield queries AI engines. You can run hourly polls for critical facts, daily for general monitoring, or weekly for evergreen claims.

Creating a Schedule

Navigate to Shield → Schedules and click Create Schedule.

Step 1: Choose Facts

Select which Truth Nuggets to monitor:

  • All facts → Monitor everything
  • By category → Just Products, Leadership, Financial, etc.
  • Custom selection → Pick specific nuggets

Example: “Create schedule for Product facts only (5 nuggets: product names, pricing, features)“

Step 2: Choose Engines

Select which AI engines:

  • ChatGPT (GPT-4, 4o, 4.5)
  • Claude (3.5 Sonnet, 3.5 Haiku)
  • Gemini (Advanced, 2.0)
  • Perplexity (Pro)
  • Copilot
  • Others (coming soon)

You can monitor different engines on different schedules.

Example: “Monitor ChatGPT daily, Gemini every 6 hours, Perplexity weekly”

Step 3: Choose Frequency

Set how often cross-checks run:

FrequencyExecutionUse CaseCost (approx)
HourlyEvery 60 minutesCritical factsHigher API spend
Every 6 hours4x dailyImportant factsModerate
DailyOnce per dayStandard monitoringLow
WeeklyOnce per weekHistorical/evergreen factsVery low
MonthlyOnce per monthRare updatesMinimal
CustomYou define scheduleAdvanced usersVaries

Step 4: Choose Time (Optional)

For schedules that don’t run every hour:

  • Daily: What time? (default: 2:00 PM UTC)
  • Weekly: What day/time? (default: Monday 2:00 PM UTC)
  • Monthly: What date/time? (default: 1st of month, 2:00 PM UTC)

This prevents thundering herd (all queries at once) and lets you choose low-traffic times.

Step 5: Name & Save

Give the schedule a name:

  • “Product monitoring (hourly)”
  • “Brand facts (daily)”
  • “Competitive facts (weekly)”

Click Create. Monitoring starts immediately.

Managing Schedules

View All Schedules

Shield → Schedules shows all active schedules:

Schedule 1: Product Monitoring
├─ Facts: 5 (Shield features, pricing, availability)
├─ Engines: ChatGPT, Claude, Gemini
├─ Frequency: Hourly
├─ Created: March 10, 2026
├─ Next run: In 23 minutes
└─ Last run: Completed successfully


Schedule 2: Brand Monitoring
├─ Facts: 12 (Company, Leadership, Financial)
├─ Engines: ChatGPT, Claude
├─ Frequency: Daily (2:00 PM UTC)
└─ Next run: Tomorrow at 2:00 PM

Edit Schedule

Click any schedule to adjust:

  • Add/remove facts
  • Add/remove engines
  • Change frequency
  • Pause temporarily

Changes take effect on next run.

Duplicate Schedule

Copy a schedule to create a variant:

  • Original: “Product monitoring (hourly)” for 5 engines
  • Duplicate → “Product monitoring (daily)” for 2 engines

Saves time compared to creating from scratch.

Pause/Resume

Temporarily disable a schedule:

  • Click Pause → Stops monitoring, won’t run next scheduled time
  • Click Resume → Resumes next cycle

Useful if you’re updating Truth Nuggets or troubleshooting.

Delete Schedule

Permanently remove a schedule:

  • Cross-checks stop running
  • Historical data remains in audit log
  • Can recreate anytime

Best Practices

Start Conservative

Begin with:

  • 1-2 engines (ChatGPT + Claude cover 60%+ of usage)
  • Daily frequency (low cost, captures most issues)
  • Core facts only (company, main products, CEO)

Then expand based on results.

Tier by Importance

Critical facts (hourly):

  • Product names and core features
  • CEO/founder names
  • Key differentiators

Important facts (daily):

  • Pricing
  • Company stage (funding, size)
  • Technical specifications

Evergreen facts (weekly):

  • Founded date
  • Headquarters location
  • General industry positioning

Monitor Growth Leaders

If certain AI engines cite you frequently:

  • Monitor them more often (hourly)
  • Others can be weekly

Example: “ChatGPT mentions us frequently; monitor hourly. Perplexity rarely mentions us; check weekly.”

Seasonal Adjustments

Around major events, increase frequency:

  • Product launch: Move to hourly for all engines
  • Quarterly earnings: Daily for Financial facts
  • Fundraising: Hourly for Financial facts
  • Normal times: Daily or weekly

Then return to baseline after event.

Cost Optimization

Calculation

Cost depends on:

  • Number of schedules (independent)
  • Engines per schedule (multiplier)
  • Frequency (multiplier)
  • Facts per schedule (each query is 1 API call)

Example:

  • 3 schedules
  • Schedule 1: ChatGPT, daily → 1 query/day
  • Schedule 2: ChatGPT + Claude, daily → 2 queries/day
  • Schedule 3: Gemini, hourly → 24 queries/day
  • Total: 27 queries/day ≈ 810/month

At typical API rates: ~$1-2/month in direct costs (included in subscription).

Cost-Saving Tips

  1. Consolidate schedules → One schedule with 5 engines is cheaper than 5 schedules with 1 engine each
  2. Use daily instead of hourly → Most hallucinations detected daily; hourly rarely helps
  3. Remove low-signal facts → If a fact never triggers alerts, lower frequency or remove
  4. Bundle similar facts → Group related facts in one schedule

Cost vs Detection Trade-off

FrequencyCostDetection Latency
HourlyHighMinutes (fastest)
6-hourlyMedium-HighHours
DailyLowHours to 24h
WeeklyVery lowDays

Most organizations use daily as the sweet spot.

Monitoring Schedule Health

Go to Shield → Schedules to see health status:

  • Green = Running perfectly
  • Yellow = Minor delays (one failed run, recovered)
  • Red = Multiple failures (API issues, needs investigation)

Click any schedule to see:

  • Last 10 runs (timestamps, success/failure)
  • Failures with error messages
  • Average latency per run

API-Based Scheduling

For advanced users, configure schedules via API:

Terminal window
curl -X POST https://api.truthvouch.com/v1/schedules \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "name": "Product monitoring",
    "nugget_ids": ["nugget-1", "nugget-2"],
    "engines": ["chatgpt", "claude"],
    "frequency": "daily",
    "time": "14:00:00Z"
  }'

See Trust API docs for full reference.

Next Steps