AI data that seems inconsistent or false
The five most common causes
1. Wrong site configured for GSC / GA4 / Bing
Most common case: you created project « ChatDirect » but mistakenly configured GA4 on Armoben's property_id. Result: GA4 data shown in ChatDirect is actually Armoben's. Solution: in Projects > open ChatDirect, verify each GSC/GA4/Bing dropdown and reselect correct property. Contaminated cache is automatically purged.
2. Missing recent scan
Displayed data always comes from last consolidated scan. If your last scan is 3 months old and you changed domain meanwhile, data is obsolete. Launch new scan.
3. Natural AI variations
AIs aren't deterministic. Same prompt can produce different responses between scans (model temperature, corpus update, randomization). A ±10% variation on a metric is normal. Analyze trends over multiple scans rather than isolated scan.
4. Poorly calibrated prompts
Too-generic prompt (« best insurance ») gives different mentions between scans based on AI chance. Precise prompt (« best auto insurance for young driver in Quebec in 2026 ») gives much more stable results. If data varies wildly, review prompt wording.
5. Inconsistent filtered period
Some pages filter by default on 30 days, others on 7 or 90. If you compare two numbers on two different pages, verify period filters. Numbers may seem inconsistent when they measure different time windows.