Hey all,
Two developments made the cut this week. There was also the usual mountain of recycled SEO speculation, but life is short.
Google’s AI Search Opt-Out Is Now Available Globally
Status: Confirmed
Google completed the global rollout of its new Search Console control on 31 August. Verified site owners can now exclude their content from AI Overviews, AI Mode and generative Discover results without that decision becoming a ranking signal in conventional Search. Reuters reported the completed rollout and the ongoing EU scrutiny.
The important bit: opting out does not magically send the missing AI clicks back through the normal blue links. It simply removes your pages from Google’s generative results.
Why It Matters
For affiliate review sites, AI visibility is still a potential source of discovery, citations and - if you’re lucky - actual visitors.
Unless Google’s AI features are reproducing enough of your content to cause measurable harm, opting out may amount to responding to reduced traffic by requesting even less visibility … bold strategy!
“don’t cut off your nose to spite your face”
What To Do
Leave review sites opted in for now and record:
- AI impressions
- AI clicks
- Affiliate conversions from AI landing pages
- Queries and countries producing that visibility
If you want to test the opt-out, use an isolated low-risk property rather than the entire money site. Compare its conventional rankings before and after to verify Google’s claim that ordinary Search remains unaffected.
OpenAI Released GPT-6 Astra
Status: Confirmed
OpenAI released GPT-6 Astra through the API on 3 September. It is aimed at difficult end-to-end work involving research, coding, computer use and long-running tool-based workflows. It also adds tool search, native compaction and support for contexts exceeding one million tokens. OpenAI’s API changelog contains the migration details.
The standard API price is $10 per million input tokens and $50 per million output tokens, compared with $4 and $20 for GPT-5.6 Sol. Batch processing costs half the standard rate. Current Astra pricing is here.
(don’t tell Ed Zitron)
Why It Matters
Astra could be useful for the difficult parts of an automated review pipeline:
- Reconciling conflicting research
- Auditing claims against source material
- Repairing failed article batches
- Operating browser-based publishing tools
- Reviewing an entire project or repository in one context
It would be a rather expensive choice for keyword sorting, metadata or routine article drafting.
Using Astra to generate 500 meta descriptions is a bit like hiring Sherlock Holmes to alphabetise the spice rack.
He could do it.
There are cheaper options.
What To Test
Run Astra against Sol on ten genuinely difficult tasks - not fifty ordinary articles.
Score each result for:
- Unsupported claims
- Missed evidence
- Manual corrections
- Completion rate
- Total cost per accepted result
Keep Luna or Terra handling bulk preparation and first drafts. Use Astra only if its reduction in errors or manual work covers the additional cost.
Also note that tool calling with Astra requires the Responses API. Existing Chat Completions workflows will need migrating before using its agent capabilities.
Reddit and Parasite SEO Watch
Not too many developments last week here.
Reddit’s earlier decline in ChatGPT citations still lacks an official explanation, while Google’s different parasite-enforcement rules inside and outside the EEA are too new to produce trustworthy conclusions.
For now, keep recording the same commercial queries from the UK and Ireland. Let the data arrive before declaring parasite SEO either reborn or buried for the seventeenth time.
My Priority This Week
Do not rush Astra into the full production pipeline.
Test it as a final research and QA layer on a small batch while keeping the cheaper models underneath. At the same time, confirm that your important Search Console properties remain eligible for AI visibility.
More intelligence at the top, cheaper labour underneath.
Basically a traditional British organisation chart 
Onwards!
- Rohan