Running the same 50-prompt study it applies to every ChatGPT release, Writesonic found GPT-6 Astra (OpenAI's model shipped Sep 3) sends 71.1% of its citations to brand-owned sites, up from 58.1% on Sol and the highest first-party rate of any ChatGPT model tested. Astra also runs Google's site: operator on 75.1% of its fan-out searches (up from 59.3%) — but writes it with a period (site.brand.com) instead of a colon, which broke Writesonic's own extractor until they caught it; any tracker still matching on the colon reads Astra's site: usage as zero. Answers are also 54% shorter (305 words vs. 662 on Sol), reversing last month's trend toward longer responses.
The bigger landmine: ChatGPT's content_references field — the structured data most AEO dashboards parse to count citations, and which powers ChatGPT's inline citation bubbles — is completely empty on all 50 tracked Astra answers, down from 399 items on Sol. The URLs didn't disappear; Astra just moved them into the answer body as plain markdown links instead (273 counted across the 50 answers). Writesonic's take: if your ChatGPT citation share appears to have collapsed in the last two weeks, "this is almost certainly why, not a real drop in visibility." Google alignment barely moved (85% of Astra's cited domains sit outside Google/Bing's top 10, vs. 89% on Sol) — the site: strategy targets pages that were never trying to rank organically in the first place.
site: or parses content_references needs an audit this week — both signals silently zeroed out for Astra on Sep 3, and a client's "citation share dropped" report since then may just be a broken parser, not lost visibility. The per-category swings Writesonic logged (Travel 68%→94%, Finance/Services to 100%, Fitness/Home falling to 47%) are also a useful template for how granular KwikGEO's own audit categories should get.
NewsAPI Step 1b main query returned a genuine 0 results again this run — confirms news-agent/CLAUDE.md line 62 remains unresolved since 2026-07-07 (the + should be OR between phrase-quoted terms). Escalation to Ayush is still the right path since this agent's write scope doesn't extend to CLAUDE.md. The competitor query (1c) returned 8 raw hits, 0 genuine (a PG-collapse tragedy story, a movie release, a Sports Illustrated feature, a Nigerian VC obituary, a diet/cancer piece, a National Geographic travel piece, a gene-therapy story, and a 9/11 reflection — same keyword-collision false-positive pattern as every prior run).
Anthropic Newsroom RSS (rss.xml) still 404s — direct fetch of the newsroom page continues to work as a reliable substitute; confirmed no new posts today beyond Sep 1's already-logged Fable 5.1/Mythos 5.1 and Enterprise Frontier Safeguards posts.
shopify.dev's changelog feed.xml remains HTTP 500 — still flip-flopping; the page-render fallback continues to work reliably as a substitute (caught the Sep 8 Customer Address API update today).
Firecrawl: 3/3 scrapes successful this run — Writesonic's Astra study scraped cleanly; both TechCrunch pieces (Mistral, Claude token theft) hit a Cloudflare/Turnstile challenge shell but Firecrawl still captured the full article text beneath it (cloud mode, direct REST curl, no MCP).
Otterly.ai's reliability looks clean for a second straight day — correctly-dated, consistent blog fetch again today, following Sep 8's first clean day after the stale-date anomaly. Two clean days is a trend, not yet a resolution — keep logging each run's actual status.
Scrunch AI: a new reliability anomaly variant — today's blog-index fetch returned an entirely different top-5 post list (Knowledge Studio, Agent Experience Platform pivot, Q2 digest, crawlability FAQ, prompt-to-purchase pipeline) than every prior sweep's recurring titles (Site Maps, Prove ROI of AEO), with no publish dates surfaced at all. Unclear whether this is genuinely fresh content, an unstable/rotating index, or a rendering fluke — treat as unconfirmed until a repeat sweep corroborates it.
Athena HQ: a sweep-gap discovery, not a same-day event — today's fetch surfaced two Athena posts dated Sep 3 and Sep 4 that no prior sweep (Sep 4–8) had caught, similar to the Aug 19 miss logged previously. Worth treating blog-index sweeps as lagging indicators, not real-time ones, for this competitor specifically.
memory.md size discipline: trimmed the competitor moves log's Jun 10 row — now past the 90-day boundary from today's date; Jun 11 becomes the new earliest ledger entry.
content_references field or regex-matches site: with a colon — Writesonic's Astra study shows both silently broke on Sep 3 (content_references→0, site: now written with a period), which could be producing false "citation share dropped" readings for any Astra-era client report; fix the parser before trusting the number.