On 15 May 2026, Google published their first official guide to showing up in AI search - the answers in AI Overviews and AI Mode. When it landed, I posted a quick reaction: this is big, but it's Google, so read it accordingly.
In the five weeks since, a stack of vendor (read vested interests!) reports, a Google Marketing Live keynote, and a platform change to how ChatGPT links out to websites have all piled on top of it.
If you run marketing for a Christchurch/South Island business (likely with a team of two!), you cannot read all of it and still do your job. So here is the filter I use instead.
Four questions
I call it the Ridiculous Conflicted-Source Filter. Run this over any "best practice" before it gets near your hard-fought marketing and comms budget for the year:
Run it on Google's guide
Google sells search - its dominance, and the ad inventory on top. The guide's headline message is that "AEO" and "GEO" are just marketers' words, and that optimising for AI search is still SEO done well. Run the four questions over it:
The part to be wary of is the suggestion that you can stop worrying about everything else. That bit serves Google. That doesn't make the guide wrong. Much of it is sound.
The point is to notice who wrote it and read accordingly.
Now try this filter on LLMs.txt
Now run this filter on Google's advice to not add an llms.txt file to your site. The guide says that it has no meaningful benefit. Think about how not having an LLMs.txt file benefits this source, and how it ignores the rest of the AI/LLM bots?
And then, interestingly enough, Google own internal documentation confirms that their own Chrome Lighthouse algorithms use LLMs.txt?
Can you see how this is conflicting? And how it keeps you from being more AI/GEO ready for all other non-Google platforms?
Turning the filter on a source I rely on
I cited Adobe's 2026 AI traffic report in a recent LinkedIn post - AI-referred visitors converting 42% better than other channels, revenue per visit up 37%. So I have to run my own filter on it. Adobe sells experience and analytics software. Its headline finding, that a large share of retail content is invisible to AI, points you toward analytics and optimisation tooling.
By question four, the claim holds. The conversion data is first-party measurement across roughly a trillion visits, hard to fake, and matched by Similarweb (a longtime trusted source), which noted ChatGPT referrals converting at 7.1%, second only to paid search. The incentive is there, the evidence is solid, the claim survives.
This is the bit worth clarifying. That doesn't make Adobe wrong. Their data may well be sound, and probably is. The same goes for Google, or any big tech company publishing advice. Assume good faith if you like. Then notice the vested interest, and weigh the claim against it rather than taking it as gospel.
What keeps surviving the filter
Run it across enough of this material and the same few survivors keep turning up, whoever's selling what:
Why this matters more here
Most of the AI search advice reaching a South Island marketing manager was written for a US enterprise with a team to match, by someone with something to sell. Some of it is useful. Some of it is a billable line item (aka another tool) dressed up as best practice. The marketing budgets that do well out of this are the ones where the marketing manager who read every guide, this one included, asked who wrote the thing first.
GEO and AI Ops can be overwhelming, and Ridiculous Digital can help take it out of your (genuinely) too hard basket - helping you with which bits of AI search advice to act on for your brand and which to quietly skip. Just reach out.
So, a question: what is the last piece of AI search (GEO) advice you took at face value - and would it survive these four questions?