Tom Wilson played Biff in Back to the Future. Three iconic films, the most recognisable bully in cinema history and for the next 40 years a steady stream of strangers asking him the exact same handful of questions.

“What's Michael J Fox like?”. “Was that real manure?”. “C’mon look mean”. “Back to the Future 4?”.

He heard it so many times he wrote a song. It's called the Question Song and he still performs it. There's a line in it that has stuck with me.

"When I'm flying in a plane or I'm on the street, there's a lot of friendly people that I like to meet. They shake my hand and never ask my name, and they start asking questions that are always the same."

So he answered all of them. In a song. So he'd never have to do it again. He even carries a card that pre-answers the lot. Michael J Fox is nice. Crispin Glover is unusual. The hoverboards didn't fly. Done.

It is very funny (unless you’re Gary Busey) and here's why I keep thinking about it in relation to what I do in AI search.

Wilson got asked the same questions so often that he could automate the answer. Predictable input, cached output. You ask the obvious thing, you get the worn-out thing back, the same line everyone's already heard.

That is what a large language model does too.

Feed it the question everyone else is typing and you get the median answer. The safe, sanded-down, seen-it-a-thousand-times response. Same prompt, same boring reply. You're not getting insight, you're getting Biff's card.

The interesting stuff only shows up when you ask something the model hasn't effectively answered a thousand times already. The question nobody else thought to ask. That's where the value is now.

Take a simple one. "Where should I go for dinner tonight?"

Ask that cold and you'll get a generic list. The same chain restaurants and safe bets anyone in any city would get back. Disappointing output.

Now plug your email in. Suddenly the model can see the confirmation for your hotel, so it knows you're in Soho not at home. Add to the prompt that you fancy small plates, you're vegetarian, you've got 45 minutes before a show and you hate a loud room, and the whole thing changes. The lazy question got a lazy answer. The detailed question, with the connectors feeding it real context, got you to a decision in one go.

Same model. Same dinner. Completely different answer. The difference was entirely in what you gave it to work with.

And this is the bit people ignore. The model is not the edge any more. Your competitor can open the same ChatGPT, the same Gemini, the same Claude as you. What you ask it, and what you connect to it, is the edge.

In a world where the answers are infinite and free, the scarce skill is the asking. The people who get further won't be the ones with access to answers. Everyone has that. It'll be the ones who frame the question better than anyone else in the room.

What this looks like for a marketer

Enough theory. Here's what I actually do.

I run Fireflies on every meeting. With my co-founder, team, clients, all of it recorded and transcribed. That becomes a knowledge base I can go back to, not just to remember what was said but to work out the right questions to ask next based on the actual discussion. The answers are only as good as the context, and the context is sitting in those transcripts.

I keep it organised into projects in Claude so it's not one big mess. Structure matters. A tidy knowledge base gets you a better answer than a junk drawer of notes.

For writing detailed prompts, Wispr Flow does the heavy lifting. I've mentioned it before. It's far easier to narrate my thoughts than to type them, and because the prompt ends up longer and more detailed, the output comes back better. Talking gets more context out of your head than typing does. Highly recommended if you have lots of thoughts in your head that you struggle to get down on paper.

I also connect Notion, which is the knowledge base for both businesses. Project deliverables, contracts, the P&L, all of it. Once it's connected there's no dropping big chunks of data in or wrestling with it manually. I just ask. Treat it like a personal assistant that already knows your business inside out.

One more, and it's the easiest to start with. Ask the LLM to ask you questions before it answers. Tell it to interrogate the task until it fully understands it. AI lets you take on more and more, but the context switching is a real drain and you miss things. Getting the model to ask questions back fine-tunes the prompt and pulls out the detail you forgot to give it. The best answers often start with the model questioning you, not the other way round.

That last one brings it full circle. Better questions, in both directions.

The same thing happens in AI search

There's a GEO angle here too, because of course there is.

When someone runs a prompt through ChatGPT, the model doesn't search for your exact words. It rewrites the question first, then goes and looks, fanning one prompt out into several. We see this in the data we pull at Obsero. The query the search engine actually runs is not the query the user typed. It's been reshaped, tokens dropped, intent rebuilt.

So the framing of the question decides the answer twice over. Once for you as a user. Once for the brand trying to get surfaced inside that answer. Lazy prompt, generic result. Lazy GEO strategy, generic result. Same pattern, every time.

Tom Wilson worked this out 20 years ago, standing on a stage at 3am getting asked the same thing for the hundredth time. The questions that are always the same get the answers that are always the same.

Ask better questions.

Managing your cap table doesn’t have to be complex. Pulley simplifies equity management for Founders and CFOs with intuitive workflows, accurate, audit-ready reporting, and predictable pricing—so you can plan and scale without surprises. Onboard in days, not weeks, and rely on responsive, expert support every step of the way.  

From issuing grants to 409A valuations or ASC 718 reporting, Pulley gives you the clarity to manage equity, make decisions, and get back to work. Experience a platform built for business owners and finance teams: transparent, reliable, and designed to put your company first.

See Pulley in action

The world's top marketers will be making their way to the Côte d'Azur this week for Cannes Lions, and my Obsero co-founder Mike Logue will be amongst them.

If you'll be there and want to talk all things AI search and the future of discoverability, get in touch with Mike. The rosé is on us.

That is it for this week.

I think it was fitting on Father’s Day to reference Back to the Future.

It is my favourite movie of all time and it spans 40 years from the first time I watched it with my Dad to the first time I watched it with my sons. Sadly, Robert Zemeckis and Bob Gale didn’t manage to predict the Internet in 2015, but maybe - just maybe - AI could create Back to the Future 4?

Tom Wilson may have to rework that song.

Andy