The new marketing war isn’t for clicks – it’s for memory

Marketing used to be a game of visibility.

You researched the market, spotted the gaps, built the content, ran the ads. You showed up, people saw you, some of them clicked-and, crucially, you could see it happening.

You had dashboards. Attribution models. Funnels. You could test, optimize, improve. It wasn’t perfect, but it was observable.

That era is over.

We’re entering a fundamentally different paradigm; where intelligent systems mediate how people discover, evaluate, and interact with brands. Whether it’s an LLM summarizing answers, an AI assistant making recommendations, or a multi-modal agent facilitating purchases, the new digital experience isn’t about users browsing websites. It’s about systems making decisions.

And those systems don’t behave like users.

• They don’t search with discrete keywords. They don’t follow funnels. They don’t click.
• They see differently. They reason differently. They decide differently.

This shift doesn’t just break SEO or attribution.

It breaks marketing as a measurable discipline.

When an AI assistant decides what to recommend based on Reddit sentiment, embedded documentation, third-party schema, or the tone of a YouTube review, your analytics stack won’t capture it. You won’t know what tipped the balance. You won’t even know you were in the running, let alone that you lost.

In response, the industry is scrambling to claw back visibility. A wave of AI tracking tools promise insight into where and how your brand appears in generative results. Platforms offer new impression metrics – buried, blended, and unsegmentable.

But these are comfort metrics. They offer the theatre of insight, not its substance. They retrofit old measurement paradigms onto systems that were never designed to be interrogated.

This article is about that shift – and what it means.

It’s about why:

• The tools we’ve built no longer serve us.
• Brand reputation now lives in the latent soup of model memory.
• Being inferred by an AI matters more than being clicked by a human.

Because when everything is opaque, the job isn’t to be seen.

It’s to be unignorable.

The world we lost

For the better part of two decades, marketing was a game of measurable influence.

You could map a funnel. Model user journeys. Attribute value across channels. Sure, the tools were messy – tracking was incomplete, attribution was flawed, and platform data was often opaque. But there was at least a logic to the system. It was observable. Manipulable. Optimizable.

We could research what people wanted, through keyword volumes, search trends, social signals. We could place ourselves in their path via ads, content, listings, influencers. And we could observe what worked, through impressions, clicks, conversions, attribution models.

This was the golden era of the marketing stack:

• Research tools to measure demand.
• Campaign tools to reach audiences.
• Analytics tools to track behavior.
• Optimization tools to iterate and improve.

And, crucially, everything was wired into a loop. You did something. Something happened. You measured it. You adapted.

Even if you were wrong, you could know you were wrong.

That feedback loop didn’t just drive performance. It defined what marketing was. It gave us a sense of control. Of cause and effect.

Yes, the tools were flawed. Attribution was messy. Data was noisy. And in recent years, even that messiness has deteriorated further – between rising consent barriers, cookie blockers, and browsers systematically undermining tracking technologies, most of our reporting is already built on sand.

But the game was still legible. You could observe, infer, improve.

Now, the rules are changing – and we’re not the ones doing the observing anymore.

Today, users aren’t just exploring the web. Increasingly, they’re letting systems do it for them.

Instead of scanning search results or comparing reviews, people are beginning to delegate discovery and decision-making to intermediaries: LLMs, smart assistants, summarizing interfaces. These systems take a question, evaluate the landscape, and offer a single output – an answer, a product, a decision.

We can already see this shift in how Google presents information: AI Overviews distil SERPs into synthetic summaries, collapsing the decision space into a handful of unclickable takeaways. Other platforms are heading the same way.

These systems don’t behave like users. They don’t:

• Search in discrete, one-shot queries.
• Necessarily visit your site or engage with your funnel.
• Click.
• Convert.
• Show up in your data.

Instead, these systems:

• Evaluate.
• Summarize.
• Decide.

You’re not optimising for a visitor anymore.

You’re trying to be included in a model’s reasoning.

What used to be a competition for visibility is becoming a battle for inclusion.

We used to publish content to earn clicks. Now we publish signals in the hope of being cited, referenced, or trusted by a system that never tells us how it thinks.

That shift severs the connection between what we create and what we can see. It leaves us without feedback, without attribution, without certainty.

Measurement is dead

Now, marketers are flying blind.

In a world where intelligent systems shape outcomes invisibly, our tools for measuring success no longer apply.

Clicks, sessions, conversions – they might still happen, but they don’t explain why. The causal link between action and result has evaporated.

A user might never visit your site. They might act on a product recommendation surfaced in a generative answer, trust a quote pulled from a Reddit comment, or be swayed by a brand comparison made in passing by an AI assistant. They might hear your name in a podcast transcript, see it mentioned in a blog post, or catch it in a YouTube short – all without attribution, interaction, or awareness.

Every one of those moments is mediated by a system that evaluates your brand, weighs its context, and acts – without ever touching your funnel.

And yes, that conversion will show up in your reports just like any other. You’ll see the outcome, but none of the story behind it. The diagnostic power is gone.

This isn’t a minor loss of precision. It’s a severing of the feedback loop.

And that’s not just inconvenient – it’s existential. Marketing as a discipline depends on knowing what caused what. Without that, we’re guessing.

That’s why the industry is scrambling for something – anything – to measure. And so a wave of AI visibility tools now promise to show how your brand appears in generative answers. But they’re built on sand:

• Panel-based tools rely on partial data from browser extensions or synthetic SERPs. They only capture a sliver of behavior – often from skewed cohorts.
• Prompt-based tools fire zero-shot queries like “What’s the best [X]?” into LLMs and log the results. But that’s not how people use these systems. Real interactions are iterative, contextual, memory-driven. A single snapshot tells you almost nothing.
• Even Google’s own data on impressions and clicks from AI Overviews offers no useful segmentation. You can’t isolate it. You can’t interrogate it. It’s visibility without clarity.

These aren’t insights. They’re artefacts of an old paradigm, duct-taped onto a new one.

They make us feel like we’re still in control. But they don’t help us make better decisions.

They’re theatre. Comfort metrics. Shadows on the wall.

To be clear, some proxies might still be useful, as long as we treat them solely as directional. The danger is in pretending they’re precise. Because they’re not. And the systems we’re dealing with were never designed to be measurable in the first place.

The longer we cling to legacy visibility, the more blind we become to the real shift underway.

And the less prepared we’ll be for the new game: one where visibility doesn’t matter.

Only influence does.

Influence is indirect and fragile

In this new world, you don’t get to make your case directly.

You don’t persuade the user. You don’t pitch the benefits. You don’t close the deal.

You present your evidence to a system, and hope it survives the summarization.

These systems don’t rank content. They assemble consensus. They don’t reward clickbait. They reward confidence. They weigh sources, distil perspectives, and deliver answers.

And in that distillation, much of what makes you special can be lost.

You’re not influencing decisions. You’re influencing reasoning.

That reasoning is probabilistic, synthetic, and unsentimental. Models don’t just look at your best content – they average across everything they can find. Documentation. Mentions. Citations. Schemas. Reviews. Context.

You might have the best product. The smartest team. The deepest expertise. But if the system doesn’t see that – if it isn’t consistently reinforced across credible surfaces – it doesn’t count.

This makes your position fundamentally fragile.

A competitor with better documentation, cleaner markup, tighter semantic alignment, and more coherent citations may become the default. Not because they’re better, but because they’re easier to summarize.

Worse, your absence can be caused by noise. A legacy schema conflict. A lukewarm review. A forum post from 2018. A few garbled citations. None of them matter on their own. But collectively, they push you below the threshold of inference.

You won’t get a warning. You won’t see a drop in rankings. You’ll just stop being included.

The system didn’t reject you. It just didn’t think of you.

That’s the new threat – not being disliked, but being omitted.

Not failing to win. Failing to be remembered.

The reputation war

If inclusion is everything, then reputation is infrastructure.

Not brand awareness. Not vanity metrics. But the ambient, structural signal your brand leaves in the latent layers of the web – and the models trained on it.

This isn’t just about who’s talking about you. It’s about what the system thinks they mean when they do.

Your reputation is the residue of context.

The echo of consistency across the network.

And it’s shaped not just by what you say, but by what others repeat – and what the model remembers.

This includes:

• How consistently you’re described.
• Where you’re mentioned.
• Who links to you.
• What context surrounds you.

A better competitor doesn’t need to out-market you. They just need to show up more coherently. Across more surfaces. In more trustworthy formats. With fewer contradictions, more citations, and cleaner associations.

That’s all it takes to win inclusion.

But this also opens the door to sabotage.

Your visibility can be undermined (deliberately or not) by:

• Astroturfed reviews.
• Low-quality forum chatter.
• Negative sentiment loops.
• Misattributed citations.
• Fabricated content designed to erode confidence.

None of it needs to go viral. It just needs to exist, to be absorbed into the substrate.

And once it’s in, it’s hard to get out. The model won’t notify you. You won’t see a drop in rankings. You’ll just notice fewer mentions. Less traffic. Diminished presence. And no explanation.

The system didn’t attack you. It’s not biased or unfair. There is no social contract. It just learned to prefer someone else.

That’s the new competitive landscape:

• Not fighting for attention.
• Fighting for inference.

And winning that fight means not just defending your signals – but reinforcing them.

Because if reputation is the infrastructure of inclusion, then what you build next has to be unshakeable.

Marketing in the latent web

If reputation is what earns inclusion, then marketing’s role is to make that inclusion inevitable.

Not by shouting louder. Not by chasing clicks. But by becoming structurally obvious – woven into the fabric of the web in ways that models can’t ignore.

This isn’t about performance metrics. It’s about contextual permanence. You’re not just trying to rank. You’re trying to be the most consistent, confident, and computationally legible answer to the unspoken prompt: “Who belongs here?”

That means showing up everywhere that matters – and always in the same voice. It means repeating your positioning until it becomes a default assumption. And it means thinking less like a brand campaign, and more like a data layer.

You become influential not through exposure, but through consistency. Redundancy. Contextual gravity.

You’re not being seen.

You’re being remembered.

For years, we were told to “write for humans.” And that advice still matters – because human behaviour influences culture, and culture shapes the web.

But in a model-mediated world, that’s only half the picture.

These systems don’t just observe behavior. They interpret structure. They weigh phrasing, associations, citations, coherence. They draw conclusions from how information is embedded, not just how it’s received.

That means we need content that persuades humans, yes, but also content that reinforces models. Not spam. Not trickery. But clarity. Consistency. Strategic redundancy.

Not topical authority for humans – but for machines.

That means investing in four dimensions of latent influence:

• Presence: Are you cited in places machines consider credible? Docs, forums, schema, transcripts, FAQs, datasets  –  not just blogs and landing pages.
• Positioning: Are you described consistently and clearly across those sources? Or do you show up fragmented, contradictory, or vague?
• Perception: What’s the quality and sentiment of your adjacents? Are you next to trusted voices, or surrounded by spam?
• Permanence: Are your signals stable, persistent, and embedded in surfaces likely to be crawled, trained, and referenced long-term?

This isn’t SEO with extra steps. It’s branding for an audience that reads everything and forgets nothing.

You’re not just trying to win a SERP. You’re trying to shape a model’s memory.

That doesn’t show up in your dashboards. There are no conversion graphs. But the outcome is real and existential.

Because in this new ecosystem, inclusion isn’t earned through engagement.

It’s earned through embeddedness.

Defensive brand strategies

In a world where you can be forgotten without warning, you need to build your brand like it’s under siege – because it is.

Defensive strategy means treating reputation not as a campaign, but as a system. Not something you broadcast, but something you maintain.

You’re not building a monument.

You’re tending an ecosystem.

That means building redundancy across every axis:

• Multiple trusted voices saying the same thing.
• Multiple surfaces reinforcing your positioning.
• Multiple third parties confirming your value.

If one citation gets diluted or poisoned, another can backfill. If one channel falters, others carry the weight. You’re not relying on a homepage and a handful of articles. You’re establishing a semantic fortress.

It also means proactive maintenance:

• Audit model perception: Periodically prompt LLMs with naturalistic, multi-turn queries; not to generate metrics, but to spot-check how you’re being described, omitted, or misrepresented. This isn’t measurement; it’s sanity-checking inference.
• Reinforce weak signals: Identify outdated or low-trust sources and upgrade them. Refresh your documentation. Reclaim neglected listings. Submit corrections where needed.
• Patch reputation vulnerabilities: Identify and fix inconsistencies in how your brand is described across surfaces. Clarify ambiguities before they lead to confusion. Address negative discourse early, before it spreads or hardens. And don’t just respond to criticism – surround it with signals that restore confidence.

And most of all: own your category. Define it. Populate it. Be the default.

If you don’t control the frame, the model will borrow someone else’s.

You’re not just defending brand equity.

You’re defending inclusion.

The upstream advantage

In this landscape, traditional brand-building isn’t obsolete – it’s foundational.

Advertising, PR, sponsorships, cultural presence: these aren’t just tools for awareness. They’re how you generate the raw material that machines use to form opinions.

These efforts shape the conversations that shape the web – the Reddit threads, the YouTube reviews, the podcast mentions, the Wikipedia edits. They seed the citations, the sentiment, the context. They build the substrate.

The outputs might not be directly measurable in clicks or conversions, but their effects are trackable in the right frame:

• Holdout testing across geographies or cohorts.
• Brand lift studies.
• Surveys on salience, recall, and preference.
• Share of voice tracking across platforms.

The key is shifting the mindset – from deterministic attribution to probabilistic impact. From “this ad caused this sale” to “this campaign lifted salience by X% in this cohort.”

It’s not as satisfying as watching real-time dashboards tick up. But it’s more honest. And in this new world, it’s how we stay grounded in reality – without pretending to see through the fog.

As Barry Adams, renowned SEO expert and founder, Polemic Digital, put it:

• “Marketers have been clinging to the illusion of measurement. Analytics platforms have provided the mirage of attribution and accountability, but that has always been a lie. What AI does is remove the mostly fake data that powered this illusion. It’s back to pre-internet basics, which is probably better for the marketing discipline as a whole.”

None of this is new. But in the context of model-mediated experiences, it becomes newly urgent. Because the real value isn’t just in what these campaigns do today – it’s in what they cause to be said tomorrow.

If you want to shape the model’s outputs, you must shape the culture that trains it.

In a world where direct influence is disappearing, upstream influence is your leverage.

In a world without signals, build for the shadows

This isn’t a passing trend. It’s a phase shift.

The tools we relied on – dashboards, attribution models, keyword lists – were never the point. They were artefacts of an era where visibility was linear and trackable. That era is ending.

What comes next isn’t about tactics. It’s about posture.

Because if you can’t track what works, you have to build what lasts. If you can’t prove your impact, you have to design for belief.

The web is still full of signals. But they’re no longer for us. They’re for the machines. And the machines aren’t watching for clicks. They’re absorbing structure.

So shape the discourse. Harden your reputation. Reinforce your adjacents. Leave traces in places that models crawl, weigh, and remember.

Not because it’s measurable.

Because it’s necessary.

You won’t get a report telling you it’s working. You won’t see the moment you’re forgotten.

But that’s the risk. Not rejection. Omission.

Because in this new world, it’s not enough to be the best. You have to be undeniable.

And that doesn’t start with visibility. It starts with being remembered.

This article was originally published on Jono Alderson’s website (as Everything is now opaque) and is republished with permission.