Hey all,
I’ve been trawling through the latest SEO news, Google announcements and the usual Reddit threads where everyone’s traffic has either doubled or vanished overnight.
Sometimes both.
Here are four developments worth watching if you run affiliate sites or automate content production.
Nobody needs to set fire to their website just yet 
Google Rankings Are Bouncing Around Again
There has been significant ranking volatility since the beginning of August.
Breaking news: Google has just confirmed an August update. Its official ranking dashboard now lists the August 2026 spam update as the latest confirmed event.
It started today so maybe the recent volatility was a little warm-up party … lol.
The recent turbulence now seems confirmed.
That matters because traffic can now move for several reasons:
- Traditional rankings
- AI Overviews
- AI Mode
- Discover
- Video results
- Google moving the furniture around for the sheer joy of it
I wouldn’t rewrite half a site based on a few strange days in Search Console.
What I’d Test
Choose 30 URLs:
- Ten standard review pages
- Ten video-led pages
- Ten mass-produced supporting articles
Record rankings, indexation, clicks and AI visibility each week. Leave those pages untouched for four weeks so they act as a control group.
Otherwise, you change twelve things at once and confidently declare that number seven fixed it.
Classic SEO.
Search Console Is Revealing More AI and Video Data
Google’s new generative AI reports can identify pages receiving impressions through AI Overviews, AI Mode and generative features in Discover.
Google has also made platform properties globally available for YouTube, Instagram, TikTok and X.
That means video-led affiliate campaigns can finally measure how their social and video content performs within Google.
Nice
The more interesting bit is query fan-out.
Google confirms that AI Mode may perform several related searches before constructing its answer. One product query could trigger searches about:
- Reviews and complaints
- Ingredients or features
- Pricing and refunds
- Alternatives
- Specific problems the product claims to solve
Google explains this in its AI features documentation.
What I’d Test
Build a ten-page cluster around one product covering those different angles. Compare its AI visibility against an older cluster created through conventional keyword targeting.
I’d also connect any relevant YouTube account to Search Console and export the first 28 days of data.
Do it now before Google hides it under three menus and renames it “Helpful Content Experience Insights Plus.”
Reddit Marketing Is Becoming Less Forgiving
A recent investigation by The Verge examined agencies using AI-written posts, bots and sock-puppet accounts to influence Reddit discussions and AI answers.
Moderators are responding with stronger filters, automatic bans, thread removal and brand-level keyword blocking.
Google subsequently said Reddit receives “no special preference” in traditional rankings or AI Search.
That was a spokesperson statement rather than formal ranking documentation. It doesn’t prove Reddit has no practical advantage. It just means Google denies having a large button marked:
Put Reddit above the affiliate sites.
Google’s spam policy now also covers attempts to manipulate generative AI answers.
A convincing persona won’t protect a campaign if every conversation mysteriously leads back to the same product. Moderators look at patterns.
What I’d Measure
For every Reddit campaign, track:
- Is the contribution still live after 7 and 30 days?
- Did Google index the thread?
- Does the brand appear in relevant AI answers?
Treat those as separate outcomes.
I’d concentrate on genuinely useful contributions and earn the right to mention a product. Reddit users can smell forced promotion at fifty paces.
Even better is to engage naturally within your chosen niche and behave like an expert (easy with AI … lol). Find threads where you can show off how your pro workflow creates amazing shorts, for example. Then when the time is right suggest the user DMs you for the rec … of course, subsequent visitors will see this and also DM you - mwah ha ha ha
Stop Using the Most Expensive AI for Everything
OpenAI’s current model range gives us three useful production levels:
- GPT-5.6 Luna for cheap, high-volume processing
- GPT-5.6 Terra for normal production work
- GPT-5.6 Sol for complex research and reasoning
The current model documentation and API pricing make Luna particularly interesting for bulk workflows.
I’d route tasks like this:
Luna
- Keyword classification
- Metadata
- Title cleanup
- Duplicate detection
- CSV validation
Terra
- Outlines
- First drafts
- Style-controlled rewrites
- Semantic coverage
Sol
- Research reconciliation
- Disputed claims
- Final QA on important commercial pages
You could test this on 50 articles and measure cost, factual corrections, rejected drafts and manual editing time. If Luna performs a task perfectly well, there’s no reason to pay Sol to do it.
That’s like using a Formula One car to go to the shops.
Impressive, but the neighbours will talk.
My Main Takeaway
Before producing another mountain of content, improve the measurement.
This week (if I had the time) I’d:
- Connect relevant YouTube accounts to Search Console
- Establish a fixed control group of URLs
- Track AI visibility separately from ordinary clicks
A page can gain visibility inside AI Mode without showing much conventional ranking movement. It can also receive thousands of AI impressions without producing useful traffic.
That’s the awkward bit.
Google can show our content more often while sending us fewer visitors.
For the company that proclaimed it must “do no evil” it’s quite an evil little achievement when you think about it.
Anyway … those are the developments this week. I’m going to try and make this a regular thing 
we’ll see!
Always look at what your own tests actually show because the SERPs usually tell a slightly different story from the official announcement
Onwards!
- Rohan
