AI adoption in business, deepfakes and GEO: the October 2026 wrap-up


Deepfakes and fake identities: The AI influencer who raised $12M and never existed
On September 18, 2026, someone uploaded Jean Phil's first video to Instagram. A blond guy shadow-boxing in the street, without saying a single word. No music. No text. In nine days he had a million followers. In sixteen, 230,000 on a single platform and 75 million views. Nobody asked who he was.
The answer: Jean Phil does not exist. An AI-generated character operated from a Belgian account. Before the reveal, the account had linked to $JEANPHIL, a meme coin on Solana. In under three weeks the token hit $12 million in market cap. By early October it had collapsed 75%. A Cornell University study that same month documented that users distinguish AI from human content with accuracy barely above chance.
AI adoption: Why Generative AI pilots shine in the demo and change nothing on monday
Over the past few weeks I have sat down with a financial infrastructure company, a cooperative, a foundation and a team of 80 people already using AI every day. At the table, the same story: "we did the training and nothing moved." McKinsey's 2026 State of AI report: 78% of organisations that implemented generative AI reported no measurable business impact after six months.
I have been a DJ since 1998. Nobody learns to mix by reading. You practise, the kick drums fall out of sync, you correct, you try again. With AI it is exactly the same: what delivers results is using it with real work, making mistakes in front of your team. The gap is between the world of AI training, where KPIs are certificates issued, and the world of operations, where the KPI is whether Tuesday's work goes faster. That is why we built audacIA. The practical AI adoption approach for corporate teams, SMEs and nonprofits.
Think big. Start small.
How Reframing AI's Purpose Unlocked Real Value for an EdTech Platform
An education sector client brought us in to analyse their new AI-powered learning platform. The hypothesis seemed airtight. When we started working with real users, it collapsed in three days. In our survey, 84% of students said personalised feedback from their teacher was one of the clearest signals they were making progress. Nobody wanted to replace any part of the in-person experience.
But students were using the AI platform enthusiastically outside the classroom, to practise in private, without the social pressure of raising their hand. When they arrived in class they were more prepared. The platform was not competing with the teacher: it was building the confidence to engage with them more. When we repositioned around that insight, everything clicked.
How to Write for SEO and AI at the same time without losing your human voice
In October, Google AI Overviews surpassed one billion active users. Perplexity grew 400% in user base over the past twelve months. ChatGPT Search now indexes web pages in real time. Most content published today is optimised for 2019 SEO. An AI answering questions looks for attributable sources, direct statements, structure that lets it extract specific answers. Content not built that way simply does not appear in generated responses.
This month we started applying GEO and AEO across several client projects. Measurable gains in AI Overviews visibility typically appear within four to six weeks. The architecture an AI needs is the same one that makes content clearer for people: descriptive headings, attributable statements, sources linked to the original, a structured FAQ at the end. Classic SEO and AI visibility are complementary, not competing.
What a soap crisis reveals about how teams really perform under pressure
This month I wrapped an event of over a thousand people that we have been running for five years: four days, fifty-four countries, dozens of speakers, very few hours of sleep. This time it was hand soap. "Kesiah, I just came from the bathrooms and there is no soap, what do we do?" I had to dash to the nearest supermarket, walkie-talkie in hand, still coordinating everything else. Always smiling.
Operational excellence is not what happens when the plan goes perfectly. It is what happens when the plan falls apart and the team knows exactly what to do without anyone telling them. After every event we run a complete debrief where we go over everything that happened. This is where real resilience is built. This year we are adding: reserve stock of hand soap in the warehouse.
What is worth reading and listening to this month
Four pieces the ideafoster team has been sharing privately this month. No AI-generated summaries: these we actually read.
What people ask us most about this month's topics
What is the AI pilot purgatory and why does it happen in so many companies?
The AI pilot purgatory is what happens when a generative AI pilot works perfectly in the demo but changes nothing in real work. McKinsey's data: 78% of these projects see no measurable impact after six months. Pilots are designed to impress in presentations, not to survive actual workflows.
How can companies protect themselves against deepfakes and AI-generated fake profiles?
The Jean Phil case shows that building a fake persona is cheap and scalable. Cornell University documented that users distinguish AI from human content with accuracy barely above chance. The answer: verifiable trust layers, real human presence, auditable history and cross-platform references.
What is GEO and how is it different from traditional SEO?
GEO (Generative Engine Optimization) is structuring content so AI search engines, Google AI Overviews, ChatGPT Search, Perplexity, choose it as a source in generated responses. Traditional SEO positions pages in blue-link results. GEO positions you in AI answers. With AI Overviews surpassing one billion users, both are present needs.
How do you write content that works for both human readers and AI search engines?
You do not have to choose. Both goals share the same foundation: descriptive headings, direct statements attributable to real people or sources, links to the original sources of cited data, verifiable authorship, and a structured FAQ. When those elements are present, the text works for both readers.



Comments