Hello again
Today is less of an “Oh, that’s interesting” message and more of a “here’s something we’ve built that can help you” one. And I think it’s one worth spending a moment with, if only to get a peep into a possible future.
The formula used to be simple...
Budget + SEO = page one of Google.
Then ChatGPT, Perplexity and Claude arrived, and people stopped searching and started asking.
So here’s the question worth sitting with:
When someone asks their AI assistant who to hire, does it name you?
And if not, why not?
Walk with me a moment
(I hear that on series all the time and wanted to use it at least once.)
Your AI assistant (some people have named theirs, no judgement) knows whatever you’ve told it. So when you say “my hamstring hurts, who should I see?”, it might remember you played padel twice this week after a month off, and that you prefer personal appointments late on Thursday afternoons. Off it goes to find a sports physio nearby, open late on Thursdays, with a reputation for padel injuries.
Can you see the problem it has, though?
Where does it find trustworthy information to make that match?
Mostly in the places Google looked: websites, ads, Facebook posts, directory listings. Thin, written to sell, and largely interchangeable.
Not much of an upgrade!
(“just right for not much of a donkey.”)
Now you (the sports physio) could ask AI to rewrite your website for the new rules (GEO, AEO, pick your trendy acronym). But then you’re feeding AI something AI wrote. It sounds synthetic because it is, and it sounds like every other site that did the same thing last month.
What an assistant wants (what it really, really wants) is what a good referral has always been. Someone credible, in their own words, saying you’re good at this.
If you have a podcast, you already have hours of exactly that. Your expertise, explained at length and unscripted. Your guests, often saying generous things about your work without being asked.
The trouble is it lives as audio... and an assistant can’t quote a sound file.
So we built Undercurrent
We take your back catalogue and:
pull out what was said, word for word
attribute every quote to the person who said it
link each one to the exact second in the audio
organise it by theme and by the questions a buyer actually asks
publish it on a public subdomain of your own site, laid out for AI assistants to read
The quotes are never created or edited by AI. Summaries and topic labels are, and the site says so.
That’s the point: an assistant can check every claim against the moment it was said.
Here’s one that’s already live…
Petro du Pisani has hosted On Change since November 2017. And produced forty-eight episodes on change, leadership, burnout and taking conscious breaks from work. Petro is a coach.
Have a look at context.onchange.co.za and pick any episode. Take #038, with Mike Stopforth, on the four-year reinvention that followed selling his agency. The page lists the questions the episode answers, phrased the way people ask them: How do I stop overthinking my next career move? Every answer links to the moment Mike said it.
Then there’s this, from Mike at 14:57: “And the moment I did ask for help, it was abundant.” And from Petro: “I’m a great proponent of using coaches for whatever you want to do.”
Now picture someone telling their assistant they’ve just left a corporate job, feel a bit lost, and want a coach in Joburg who understands sabbaticals. Which would you rather it had read: a services page, or eight years of Petro’s conversations with people who’ve been there?
Before we published a word, we asked six AI assistants (Claude, ChatGPT, Gemini, Perplexity, Kimi and xAI) the same 37 questions. Some named Petro and On Change directly, to test what they know. Most didn’t, to test whether she comes up when a buyer asks who to hire. That’s our baseline, and we’ll run the same questions again once the assistants have had time to read what we’ve published.
Caveat: It’s early days for measuring what assistants do with it. Nobody controls what an AI recommends, and anyone promising you the top spot is selling something (well, more than I am).
What Undercurrent does is give assistants a reason to pick you, and real human evidence, in context, that they can check and pass along for the buyer to hear for themselves.
When a buyer asks AI who to hire, it should be able to name you - and quote your guests saying why.
We’re opening the second round later this month. If you’d like to be one of the three to go live, mail us now.
No podcast yet? That’s a conversation we’re very good at. Mail us for that too ;-)
Let’s have an adventure!
