@Felunee - looser query borders (3 levels)
Same gates as production (300k / 30d + 150k / 15d). ONLY the GPT query-generation borders were loosened. Compare the Competition tab.
Baseline (tight) repeatedly surfaced the same ~5 cat channels. These 3 loosen how ChatGPT writes the search queries - wider net, same world, no drift. Low-relevance results are still culled by the score gate, so each report shows ~7-8 qualified channels.
Level 1 - Mild
Wider sub-angles, still inside the niche →
temp 0.7, 60% close / 40% wider. Wider tier = different framings/sub-topics of cat content.
New faces: Pawmise Tales, CaD Animals, Backyard Bird Mind
Level 2 - Medium
One ring out: siblings + parent category →
temp 0.8, 50/50. Wider tier may reach dog/pet psychology - same broad world.
New faces: Canine Decode, Canine Mind (dog-psych money channels)
Level 3 - Wide
Aggressive sibling/parent variety →
temp 0.9, 40/60. Most variety - also the first to drag in a little off-world noise (culled by score gate).
New faces: KindPaws (rescue-cat storytelling, 12 copies already)