Hey all,
Two developments originally made the cut this week. Then Google dropped another AI model and found a fresh place to squeeze in adverts, because apparently Monday’s briefing was looking far too settled.
Let’s crack on …
Google’s September Spam Update Has Finished
Status: Confirmed
Google’s September spam update completed on 8 October, having run globally across all languages since 24 September.
Google Search Status Dashboard
Why It Matters
We can finally analyse the damage without comparing sites while Google is still moving the furniture around behind the scenes.
The useful question is not simply whether traffic went up or down. It is which types of pages moved:
- Main commercial reviews
- Mass-produced supporting content
- Recently refreshed articles
- Video-led pages
- Parasite and third-party assets
If generic supporting pages took a hiding while reviews containing original evidence held steady, that tells us rather more than a massive red line on some SEO bod’s dashboard.
What I’d Do
I’d actually give it another week or so for the dust to settle but if you can’t control yourself then create a clean post-update comparison beginning 9 October.
Compare each page group against a similar period before 24 September, excluding the rollout itself. Track rankings, impressions, clicks, indexation and affiliate conversions.
Don’t delete all those articles your AI slaved over because Tuesday looked a bit ropey.
That’s how a temporary ranking problem becomes a permanent editorial cock-up.
OpenAI Released the Decisions API
Status: Confirmed public beta
OpenAI released its new Decisions API on 6 October.
It uses GPT-6 Luna to convert text and images into typed, structured answers. OpenAI says it can do this up to ten times faster than the Responses API.
This could be useful for the less glamorous bits of affiliate publishing:
- Classifying keywords and products
- Extracting claims from sales pages
- Reading labels or screenshots
- Flagging missing fields
- Sorting articles into templates
- Deciding which items need human review
Basically, all the repetitive jobs we previously made an expensive AI model ponder like a Victorian philosopher.
Why It Matters
A publishing pipeline doesn’t always need another beautifully written paragraph.
Quite often, it needs a dependable answer to something dull:
Does this page contain a refund policy, yes or no?
Using a full reasoning workflow for that is like calling an emergency meeting to decide who’s emptying the dishwasher. A lot of serious faces, several cups of tea and absolutely no need for any of it.
If I Had More Time I’d Test …
Running 100 existing product records through the Decisions API and my current workflow.
Measure:
- Correct classifications
- Missed or invented claims
- Processing time
- Cost per accepted record
- Number sent for manual review
Set a confidence threshold and route uncertain results to a stronger model or an actual human being.
It is still a beta. “Ten times faster” is slightly less impressive if it reaches the wrong conclusion before you’ve finished your brew.
Reddit and Parasite SEO Watch
Reddit remains highly visible in Google with many more AI citations, while the opportunistic PDF exploits seem to have taken a hit. Authority piggybacking still dominates for competitive health product review queries.
The Google spam update has missed a big chunk of its target when reputation is in the mix!
Continue tracking survival time, indexation, rankings and conversions separately. Treat ‘exploit’ placements as volatile observations, not foundations for the retirement fund.
Yes, it might be a short-term money machine but it’s not a real business.
Late-Breaking Google Bits
Since I first wrote this, Google has launched Gemini 4 Argon inside AI Mode for Pro and Ultra subscribers.
It is designed for more complicated, multi-step research, so citation choices could change for detailed comparisons and commercial queries.
Google’s Gemini 4 Argon announcement
Marvellous … they can’t stop moving that cheese.
Google is also testing prescription-drug adverts inside AI Mode and AI Overviews in the US. This is a limited advertising test rather than a new organic-ranking policy, but health publishers may now have sponsored results competing for the scraps.
Adweek’s report on the pharmaceutical-ad test
Google summarising our content while selling an advert above it.
Efficient, you’ve got to give them that 
My Priority This Week
Wait some more 
Then segment the completed spam update before changing content. I may get to test the Decisions API on one narrow classification job. We’ll see.
I’m also testing some other parasite platforms … not taking my own advice, yeah I know!
What can I say, the black-hat force is strong with this one
Onwards!
- Rohan