Why the Core Principles of Marketing Haven't Changed in the Age of AI

1. Blog
2. Strategy & Positioning

By Saroj Jha · March 24, 2026 · 9 min read

AI made marketing output cheap, so output stopped being an advantage. Buyers still shortlist the brands they remember and trust proof; here is which decisions to keep and which work to hand to AI.

From this article

If the human insight is weak, AI multiplies the noise.

AI has made the making of marketing cheap. In McKinsey’s 2026 survey, nearly nine in ten respondents said their organizations use AI regularly in at least one business function^[1], and about half of the new English-language articles published online are mainly written by AI^[2]. When every competitor has the same tools, the tools stop being an advantage.

What still decides results is the part AI cannot supply for you: which customers you choose to serve, what you promise them, whether they remember you when they are ready to buy, and whether they believe you. None of those principles is new, and the evidence since AI tools arrived says they still hold. This post takes the five that matter most to a founder, shows the current evidence for each, and sets out which work to keep and which to hand to AI.

AI lowered the cost of producing marketing, not the cost of being chosen. Buyers still pick from the few brands they remember, still rely on proof, and still respond to a promise made to people like them. Make those decisions yourself, from customer evidence, then use AI to carry them out faster.

What AI changed, and what it didn’t

AI changed the cost and speed of production and the way buyers research. It did not change how they choose. In 6sense’s 2025 survey of nearly 4,000 B2B buyers, 94% were using large language models, but that “has not changed their reliance on vendor content or third-party experts.” The vendor that won was already on the buyer’s Day One shortlist 95% of the time, up from 85% a year earlier^[3].

So the shortlist still forms in buyers’ heads before they speak to you, and AI research feeds that process rather than replacing it. The work that gets you onto the list is the same as before: a clear choice of customer, a promise they care about, and enough repetition that they remember it.

People still buy for the same reasons

AI has not rewritten why people buy. Customers still say yes to avoid risk or regret, to save time or money, to feel in control, to belong to a group they care about, and to move closer to the person they want to be. Seth Godin gave a chapter of This Is Marketing to the idea in its title: “People like us do things like this”^[5]. That is a fact about identity, not about algorithms.

AI can write 20 ad variations, 10 headlines and an email sequence in minutes. Every one of them still has to answer the same three questions: what tension in the customer’s life it addresses, who the customer believes they are or want to be, and which fear, frustration or desire it speaks to. If the insight behind the brief is weak, more AI only multiplies the noise.

Do this

Before you prompt anything, write one sentence for each of the three questions about your best-fit buyer, in the words a customer used. If you cannot quote a customer for each answer, go to the section on customer insight first.

Most buyers aren’t buying yet, and they buy the brands they remember

At any moment, most of your market is not shopping. Professor John Dawes of the Ehrenberg-Bass Institute estimates that because companies change key providers about once every five years, only about 5% of B2B buyers are in the market in a given quarter, and about 20% over a year^[6]. I cover what that means for positioning in Startup Positioning: How to Become the Obvious Choice.

When the other 95% do come to buy, they start from memory. Byron Sharp, also of the Ehrenberg-Bass Institute, calls this mental availability: “the probability that a buyer will notice, recognize and/or think of a brand in buying situations”^[7]. It is built by linking the brand to the situations in which people buy. In Jenni Romaniuk’s study of 17 business insurance products, each additional buying situation a customer linked to a brand lowered the probability of switching away by 5%, in 16 of the 17^[8].

That is why brand building still earns budget in B2B. Les Binet and Peter Field’s research for LinkedIn’s B2B Institute, using IPA effectiveness data, found that B2B businesses “grow best” when about 46% of the budget goes to brand building and 54% to activation^[9]. LinkedIn’s own summary rounds this to an even split, so treat it as a direction rather than a rule.

AI makes it easy to run 30 different messages. Memory is built the other way: the same few associations, repeated to the same buyers, in a form they recognize as yours.

Do this

List the five to ten situations in which a buyer would start looking for a product like yours (a new hire, a failed audit, a renewal date). Pick the two or three you can own, and check that every ad, post and page links your brand to at least one of them.

Segmentation, targeting and positioning are still your decisions

No amount of AI rescues a company that is unclear about three things: the distinct groups in its market (segmentation), which one it is choosing to serve now (targeting), and how it wants that group to think about it (positioning). Machine learning can find clusters in your data and predict who converts. It cannot decide which segment is right for your stage, which trade-offs you will accept to serve some customers deeply, or which position you will commit to and defend. Those are judgments about your strategy, and the Customer Segmentation Playbook walks through making them.

Running case: Example Co. Fictional

Example Co. is a seed-stage startup selling scheduling software to home-care agencies, which send caregivers to clients’ homes. In one quarter its two founders used AI tools to publish 60 blog posts and run 30 ad variants, all on the theme “AI-powered scheduling for every care agency.” Site traffic tripled. Demo requests went from 9 a month to 10, and only 2 a month came from agencies large enough to buy. Example Co. and all its numbers are illustrative.

The AI tools worked. The brief did not: it named no segment, no problem and no reason to choose Example Co. over a spreadsheet and a phone. More output only spread a vague message wider.

Do this

Write a one-page brief before any prompt: the segment you serve first, its problem in customers’ own words, the promise you make, the proof you can show, and the claims you will not make. Turn it into a positioning statement with the free Positioning Statement Generator, and give the brief, not a one-line prompt, to every AI tool you use.

Free tool

Turn your brief into one sentence

The free Positioning Statement Generator turns your segment, promise and proof into a positioning statement you can test with buyers and hand to every AI tool.

When content is cheap, sameness is the risk

Polished content is now abundant. Graphite’s analysis of 55,400 English-language articles sampled from Common Crawl found that the share mainly written by AI rose to about half by early 2025 and has stayed there; a separate Graphite study found such articles largely do not appear in Google or ChatGPT results^[2]. Being visible is easy. Being remembered and preferred is hard.

AI also pulls content toward the average. In a study published in Science Advances, 293 writers wrote short stories with or without ideas from ChatGPT. Stories written with up to five AI ideas were rated 8.1% more novel and 9.0% more useful than stories written without AI, but the AI-assisted stories were more similar to one another^[10]. Each piece got better; together they got more alike.

The defence is a point of view: a clear, sometimes uncomfortable belief about your market that a competitor’s AI would not produce by default. It answers what is broken about how things are done today, what you believe that others in your market don’t, and why your product exists because of that belief. For example, a fintech startup might hold that personal finance tools should behave like a coach, not a calculator; a sustainability brand, that sustainability must be radically convenient or it won’t scale. I explain why AI drafts drift toward the generic in Why AI-written content doesn’t sound like you.

Do this

Write your point of view as one sentence, “Most [category] does X; we believe Y, because Z,” and put it next to the headlines of your three closest competitors. If an AI tool could have written all four, rewrite yours until only you could say it.

Customer insight still comes from customers

The most useful marketing insight still comes from talking to customers, listening to sales calls, reading support tickets and watching how people use the product. Analytics shows where buyers drop off. Conversations explain why someone didn’t trust a claim, what they compared you against, and what nearly stopped them from choosing you.

It is tempting to ask an AI tool to play the customer instead. The research is not encouraging. When political scientists asked ChatGPT to answer as synthetic survey respondents, its average answers came close to those of a large national survey, but they varied less than real answers, shifted with small changes to the prompt, and changed when the same prompts were rerun three months later. The authors concluded such data is “not reliable for statistical inference”^[4].

Use AI after the conversations, where it is genuinely good: transcribing calls, pulling out the exact phrases customers use, and grouping them into themes for you to check. The method is in Running and Synthesizing Customer Interviews.

Do this

Hold ten customer conversations this month. Have AI transcribe them and list every phrase that describes the problem, then read the transcripts yourself before you accept any theme it proposes.

Strategy first, then AI: the order that works

Every new technology tempts teams to confuse tools with strategy: “we just need better prompts,” “more AI content,” “to automate the funnel.” Strategy answers who you serve, what you promise and how you are different. Tactics are how you express and deliver it. AI sits in the tactical layer; without a clear strategy, it only makes you faster at drifting. The order that works has four steps, and AI does most of its work in the third.

Running case: Example Co. Fictional

Example Co., one quarter later (fictional). The founders chose agencies with 50 to 300 caregivers that still schedule by spreadsheet and phone, interviewed 15 owners and schedulers, and heard the same problem again and again: filling a shift when a caregiver calls out at the last minute. The new promise was “Fill a call-out in minutes, not phone calls.” AI then produced 20 versions of that one message for ads, emails and pages. Total demo requests fell from 10 a month to 8, but demo requests from agencies that size rose from 2 a month to 6.

Do this

List your recurring marketing tasks and mark each one “I decide,” “AI drafts, I approve” or “AI runs.” Segment, promise, position and proof belong in the first column; if any of them sits in the third, move it back.

The real advantage

The core principles of marketing haven’t changed because people haven’t changed. Buyers still respond to clear promises and credible proof, still choose from a short list of brands they recognize and trust, and still need to see themselves in what they buy. AI has made it easier to do the wrong thing beautifully at scale, and the right thing faster than ever. Which one you get depends on the decisions you make before you open the tool.