Your Campaign Has a Shelf Life. What People Say About You Now Lasts Years

For forty years the most rigorous evidence in marketing has said roughly the same thing about paid advertising: it fades fast.

Binet and Field, working through the IPA’s databank of nearly a thousand campaigns, found that when the spend stops the awareness it built decays within about three months, and around half of brands see sales slip within a year. This is why the two of them argue for a 60/40 split and constant reinvestment, because brand memory is a fire that goes out if you stop feeding it.

Every marketer operating today has built a career inside that reality. You buy attention, you rent it for a while, and when the budget ends the meter resets.

Large language models opened a second track

And it does not behave this way at all. Most brands have not noticed they are already on it.

When someone asks ChatGPT for the best natural deodorant or the best carry-on under £150, the model does not run a campaign in its head. It reaches for what it has already concluded is true about the category, assembled from the sources it trusts, and it names two or three brands with reasons.

The behaviour that matters comes next. Having settled on that answer, it keeps giving it. Thousands of shoppers ask a version of the same question and get a version of the same names, and that holds until enough new, credible evidence accumulates to move the model’s mind — which can take months or years.

Dan Monheit at Thinkerbell calls this the infinite half-life of earned media, and it is the right name for it. A single piece of trusted proof does not decay on the old curve. It keeps recommending you long after you have stopped paying for anything.

Share of voice is becoming share of model

That turns visibility inside these engines into an asset with a very different balance sheet, and it now has a metric.

Jack Smyth and Tom Roach at Jellyfish named it share of model, the successor to share of voice: how often, how prominently and how favourably the models cite and recommend you against your competitors.

It is becoming the number that decides outcomes, because a person who asks an AI for a recommendation and receives three names very rarely goes hunting for a fourth. Being in the set is close to everything. Being outside it is close to invisible.

What actually moves share of model

The answer runs on a single principle in how these systems weigh evidence: the harder something is to fake, the more it counts.

• Your own website calling your cookies the best is worth almost nothing, because every brand’s website says that.
• A respected, independent publication saying it carries real weight, precisely because you could not have written it yourself.
• And in the productive middle — more plentiful than press and far more buildable — sits the layer most brands leave lying on the floor: hundreds of real customers, in their own words, saying they bought the thing and it was good.

That middle layer is the one a brand can actually build on purpose, and it is where the industry’s current appetite for faking things collides with a wall.

The web is filling with generated reviews and synthetic testimonials, and the models are being trained to distrust exactly that. As the fakes multiply, the value moves to proof that a real, identifiable person said a real thing, because that is the one input a generated account cannot convincingly produce. A verified customer post is about as hard to fake as earned proof gets, which is why an engine leans on it when it decides who to name.

This doesn’t retire the campaign. It changes the brief.

A stunt is no longer only a spike of impressions that dies by Thursday; the more valuable output is a moment real enough that people carry it into the corners of the internet the models read — the forums and threads and reviews where consensus quietly forms.

The unglamorous, always-on work of getting genuine customers to say genuine things about you, at scale, stops being a box marked social proof and becomes the engine of your standing inside the machines.

Build it on purpose

That last piece is the one worth building intentionally, and it is the reason BrandPay exists. It rewards your real customers with store credit for posting about you, which turns a slow trickle of organic mentions into a steady, verifiable supply of exactly the proof these engines trust: real people, saying real things, in their own words.

Every platform shift rewards moving before the position hardens. The brands seeding that proof into the models now will be the default answer in a year, while the ones that wait spend that year learning how expensive it is to unseat an answer that has already been set.

The clicks from this quarter’s campaign will be gone by next month. What your customers are saying about you — if you have given them a reason to say anything at all — will still be shaping the recommendation long after.

---

Start building your share of model. See how BrandPay works or read the AEO playbook for consumer brands.

Related reading: How BrandPay Makes Your Brand the One AI Recommends • Nothing Goes Viral by Accident • When Content Becomes Currency