Section 03 · The bubble, the hit map, the baseline
The traffic story
A mass 2025 publishing campaign built a ~95–99% informational traffic base; a June 2026 reassessment deflated it; and a 2024 baseline revealed most of the perceived "collapse" was ordinary summer.
The PAA publishing campaign
Both sites mass-published "People Also Ask" question posts in 2025. (The page-generator-pro tool generated service pages, not the PAA posts — per the owner.)
Page-class decomposition: the blog was the whole footprint
Windows: W2 = Mar 1–May 31 2026, W3 = Jun 15–Aug 11 2026. Blog URLs were 99.3% of TC clicks. Note the two charts share a story but not a scale — the commercial pages never had traffic to lose.
Dallas intent split, W2→W3: commercial −63% (4.3→1.6 clicks/day), informational −49% (80→41). The "informational fell, commercial held" hypothesis is not supported on Dallas — both fell. Brand diverged: TC "aaac" clicks fell 80→40→21 across successive windows (a real-world visibility decline); Dallas brand was flat W2→W3 at pos ~4.
35 properties: damage tracked blog exposure, not geography
Raw W2→W3 numbers below; where the R3 seasonality audit revised a site's reading, both are shown. Region-named sites averaged +2.8% vs city-named −13.9% — the region-name hypothesis for the June drop was rejected.
| Site | Blog share | Raw W2→W3 reading | After R3 seasonal correction (2024 baseline) |
|---|---|---|---|
| treasure-coast | 99.3% | blog −32% (337→229 c/d) | ~−38% vs 2024 norm; "−60% from peak" retired (peak was a 48× bubble spike) |
| dallas | — | blog −46% (534→286 c/d); "−54% from March peak" | Removed from June-event evidence — slow bleed from April; no pre-bubble baseline exists |
| houston (no 2025 PAA batch) | 60% | −51% from peak; step-downs on the same dates as Dallas/TC | True 2026 excess −4% to −26%; apparent health = real commercial base (7 c/d) + paid props |
| tampa (no 2025 PAA batch) | ~10% (≈90% commercial share) | −48% from peak / −28% | Collapses: 2026 ratio 0.85 beats 2025's 0.80; real break was Sept 2025. July calls 83% paid |
| phoenix | 97.7% | −44 c/d — #2 absolute loser; near-zero commercial organic; nobody complained | Blog-heavy: genuinely below 2024 norm |
| dayton | 88% | −51% | Blog-heavy: genuinely below 2024 norm |
| columbus | 83% | +11% (blog +4%, core +43%) — grew through the update | Counterexample stands |
| san-diego | — | blog +41% | Counterexample stands |
| austin | — | blog +45% | Counterexample stands (GA4 blind — GSC only) |
| central-ohio | — | +105% | Region-named and grew — kills region-name-as-June-cause |
| collin-county | — | flat | — |
The update killed a kind of blog content, not blogs per se. Columbus (83% blog) grew +11% while TC/phoenix/dayton collapsed. Open task: diff columbus's surviving corpus against the collapsed ones. Network aggregate: −22.5% raw, of which 89.5% of absolute loss came from /blog URLs — but blog was 60–99% of the base, so R3 flagged that stat as near-tautological (honest test on Houston: 68% of loss on 60% of base).
Seasonality: summer was always ~25% below spring
GSC's 16-month retention hid the pre-bubble year; GA4 still had it. Median summer/spring ratio across 9–10 GA4 properties (GSC and GA4 agree within 4 points where both exist):
Jun 11–25: the only discrete 2026 event
Also in the daily data, but separate from this event: an Apr 20–21 TC impressions-only step (64–70k→35–38k/day, clicks unaffected) and Dallas's slow bleed beginning in April. August steady state: TC ~200 clicks/day, Dallas ~270/day — decaying at 2024's normal rate.