Strategy

Digital marketing in the AI era: what actually changed

The operating-model read, not the trend piece. What AI answer engines and agents actually reprice, the four shifts they force, and a where-to-start playbook for teams that would rather rebuild than defend a dying model.

By Siddharth Surana, Founder, leapbuzz··13 min read·Updated

Editorial illustration on cream: many thin channel lines converging inward to a single glowing orange core node, the digital-marketing operating model reorganised around one AI core.

Bottom line

Digital marketing in the AI era is being repriced: AI answer engines and agents intercept the click and the reach that the old model was built on, so the winners rebuild around being the cited source and owning first-party signal.

  • The old operating model rented reach and monetised the click. Both are being skimmed off the top by answer engines and agents.
  • Four disciplines re-route through AI intermediaries: the funnel, content, inbound, and email lifecycle.
  • The new moats are entity authority (being cited) and owned first-party data (not renting your audience).
  • Measurement moves off last-click to first-party signal plus aggregate causal methods that survive zero-click and cookie decay.
  • Start with a diagnosis: score your exposure, then fix the widest gap first rather than rebuilding everything at once.

Digital marketing in the AI era: the old model, and why it is repricing

Digital marketing in the AI era reprices the operating model the whole discipline was built on. This is not a new channel to bolt on. For roughly twenty years the deal was simple. You won a position in a ranked list, you earned a click, and you converted the visitor on a property you controlled. Reach was cheap and getting cheaper, attention was rentable, and the click was the atomic unit of everything: the funnel, the attribution model, the content brief, the media plan.

That deal is coming apart at the seams that held it together. Two forces are pulling. AI answer engines resolve the query inside the answer, so the click that funded the model never happens. And the identifiers that made cheap reach targetable, third-party cookies and cross-site IDs, have decayed as a substrate. The tactics did not stop working overnight. Their price changed. Reach costs more and returns less, and the click you used to buy or earn is increasingly intercepted before it reaches you.

Here is the uncomfortable part for anyone who built a career on the old model. The repricing is not a temporary dislocation you can wait out. It is a structural change in who sits between you and the buyer. A generation of playbooks assumed a direct line: query, ranked result, click, your site. An AI assistant now stands in that line, reads the sources, and hands the buyer an answer. You can rage at the intermediary or become the source it quotes. This pillar argues for the second, and routes you to the four rebuilds it requires.

The thesis in one line The click-through and cheap-reach operating model is being repriced by AI answer engines and agents. The winners rebuild around two assets an intermediary cannot skim off the top: being the cited source, and owning first-party signal.

This is the hub of a cluster. Below, each of the four disciplines that ran on the old model, the funnel, content, inbound, and email lifecycle, gets a short read, and each one links to a dedicated pillar that goes deep. If you already know the shift and want the tactical craft of getting cited, our generative engine optimization playbook is the instrument this strategy points to. If you want the strategy first, keep reading.

Flowchart: click-through era items mapped by arrows to their AI-answer era equivalents.DIGITAL MARKETING IN THE AI ERAThe operating model is being repricedCLICK-THROUGH ERAAI-ANSWER ERARanked blue linkThe cited sourceCheap paid impressionsEarned citation and trustCapture the clickBe selectable by an agentThird-party trackingOwned first-party signal
The digital-marketing operating model, repriced: from the click-through era to the AI-answer era.

What AI search and agents actually change

Three mechanics do most of the work, and together they explain why the numbers on your old dashboard are starting to lie.

The first is zero-click. An AI answer engine, whether that is a chat assistant like ChatGPT, Perplexity, Gemini, or Claude, or Google AI Overviews sitting on top of a search page, generates the answer in the surface the user is already looking at. It reads across sources, synthesises, and often cites a handful of them. The user gets what they came for without visiting anyone. The influence still happened. The visit did not. If your entire measurement stack is built on visits, you have gone partially blind, and you will conclude the channel is dying when it is actually just paying out in a currency you are not counting. Our zero-click search strategy is the tactical read on this.

The second is citation. When the answer engine does credit a source, that citation is the new impression. It is a branded mention delivered at the exact moment of the buyer's question, inside a trusted answer, with the model's implicit endorsement. Winning a citation is worth more than winning a click ever was, and it behaves differently. You do not buy it. You earn it by being structured, evidenced, and trustworthy enough that the model reaches for you. That is a content and authority problem, not a media-buying one.

The third is agentic shortlisting, and it is the one most teams have not priced in yet. Increasingly the buyer does not read the answer at all. They ask an assistant to do the early work: compare the options, narrow it to a shortlist, recommend one with reasons. The assistant assembles that shortlist from the sources it trusts and can parse. If you are not in that working set, you are not on the list, and there is no retargeting pixel that reaches a buyer who never entered your funnel because an agent filtered you out before they knew you existed. Being the entity agents reach for is the new top of funnel.

The trap The failure mode is not that traffic drops. It is that traffic drops while your brand is still influencing buyers through citations you cannot see. Teams that read only the visit number cut the exact work that was still winning, then wonder why pipeline held while sessions fell. Measure the influence, not just the click.

The operating model: then versus now

The clearest way to see the repricing is to lay the old operating model next to the new one, dimension by dimension. Nothing in the left column stopped existing. Every row moved right, and the moves compound. Read this as the map for the rest of the piece: the four shifts that follow are the rows of this table, expanded.

The digital marketing operating model: then versus now. A method-level framework, not a measured claim.
Dimension Then (click-and-reach model) Now (AI-era model)
Discovery unit The ranked link on a results page The cited source inside an AI answer, or a slot on an agent's shortlist
Atomic outcome The click, and the session it started The citation and the mention, influence without a visit
Reach economics Cheap and abundant, buy more of it Repricing upward as answer engines skim demand off the top
Content job Rank for the keyword, earn the click, convert on page Be extractable and citation-ready, answer the question in place
Content shape Long pages optimised for keywords and dwell time Answer-shaped chunks with clear entity and evidence signals
Targeting substrate Third-party cookies and cross-site identifiers First-party signal you own, plus consented context
Measurement Last-click attribution and a clean visit trail First-party signal plus aggregate causal methods, plus citation tracking
Primary moat Budget and bid, outspend for share of voice Entity authority (being cited) and owned first-party data
Top of funnel Impressions bought or ranked into Being the source an assistant reaches for and shortlists
Owned channels A nice-to-have, secondary to paid and organic search The surface with the most leverage, because you control the relationship

One row deserves a flag, because it is the one that quietly breaks the most reporting. The measurement row. When the discovery unit and the atomic outcome both change, a last-click model does not merely undercount, it actively misattributes, crediting the final visit that an answer engine influenced upstream to whatever channel happened to catch the click. We come back to that in the measurement section, because getting it wrong is how teams talk themselves into defunding the work that is still winning.

The four shifts, and where each one leads

The repricing does not hit every discipline the same way, but it hits all four of the load-bearing ones. Here is the short read on each. Each has its own pillar that goes deep, because each rebuild is a project in its own right.

Shift 1

The funnel

The stages did not disappear. Each one now runs through an AI intermediary: answer engines at awareness, agentic shortlisting at consideration, assistant-led checkout creeping into conversion. The job at every stage moves from capturing a click to being the entity the AI cites and recommends.

Shift 2

Content

Content built to rank and be clicked is being out-competed by content built to be extracted and cited. The unit of value moves from the page view to the quoted passage, and the craft moves to entity authority, answer-shaped chunks, and structured evidence a model can trust and reuse.

Shift 3

Inbound

Inbound's core loop, create content, rank, capture the click, nurture the lead, gets disintermediated at the click. The answer resolves the query without the form-fill. Inbound is not dead, but its conversion path moves from being found to being cited, then being chosen off a shortlist.

Shift 4

Email and lifecycle

As paid reprices and third-party signal decays, the owned channel becomes the strongest surface you have, because no intermediary sits between you and the inbox. AI shifts it from batch-and-blast to predictive, first-party-fuelled lifecycle orchestration.

Take them in turn. The funnel shift is the one buyers feel first, because it changes who they hear from before they ever reach you. When an assistant assembles the consideration set, the awareness stage stops being a volume game and becomes a citation game. Our marketing funnel in the AI era pillar rebuilds each stage input, from awareness through retention, for a world where clicks vanish at every step.

The content shift is the engine room, because content is what answer engines actually read. A page written to rank can still be invisible to a model if it buries the answer in preamble or gives the model nothing extractable to quote. The rebuild is toward structured, evidenced, answer-first writing. Our content marketing strategy for AI search pillar is the move from traffic to citations, and it hands off to the GEO playbook for the mechanics.

The inbound shift is the one people argue about, usually by asking whether inbound is dead. It is not, but the honest answer is more uncomfortable than either the obituary or the denial. The mechanic that made inbound work, own the ranked content, capture the click, nurture the lead, is exactly the mechanic AI answers disintermediate. Our inbound marketing in the age of AI pillar is the contrarian read on what survives and what you should stop doing.

The lifecycle shift is the quiet winner. When reach reprices and your rented audience gets harder to reach, the audience you own becomes disproportionately valuable, and AI makes it smarter at the same time. Our email and lifecycle marketing in the AI era pillar covers the owned-channel renaissance and the first-party fuel that powers it.

First-party data and measurement in a cookieless AI world

Two of the shifts above rest on the same foundation, so it gets its own section. First-party data. When third-party identifiers decay and answer engines intercept the click, the data you own about your own audience stops being a hygiene item and becomes the substrate the whole model runs on. It is what powers lifecycle orchestration, what feeds the aggregate models that replace last-click, and what an intermediary cannot take from you because you collected it directly, with consent, in a relationship you own.

The strategic move is to stop renting your audience. Every logged-in behaviour, every email engagement, every self-reported "how did you hear about us" is a signal that survives the cookie and survives zero-click. Building that spine is a project, and it is the one we most often see teams skip because it is unglamorous. Our first-party data strategy is the build guide.

Measurement is where the repricing does its most expensive damage. A broken measurement model does worse than mislead. It actively defunds the right work. Last-click assumed a clean visit trail. Zero-click breaks the trail on the front end and cookie decay breaks it on the back end. What replaces it is a stack, not a single new metric.

Measurement in the AI era: what breaks, and what replaces it. Method-level framework.
The question you are asking What broke What replaces it
Which channel drove this specific conversion? Last-click, when the influencing touch was a zero-click answer First-party signal (self-reported source, logged-in path) plus modelled attribution, read as a trend not a verdict
Did this channel actually cause incremental outcomes? Correlation dressed as causation in a click-based report Incrementality tests and geo holdouts that do not need user-level tracking
How should I allocate the whole budget? Attribution stacking that double-counts and ignores offline Marketing mix modelling on aggregate data, cookie-independent by design
Are we influencing buyers we never see visit? Nothing measured it, because there was no visit Citation and brand-mention tracking across answer engines

The honest version of this stack accepts wider confidence ranges and reads trends over quarters. That is a hard sell to a team trained on a dashboard that promised a single deterministic path. It is also the only version that tells the truth once the click stops being the whole story. A directional read you can trust beats a precise number that is quietly wrong.

The five-market read

Digital marketing in the AI era plays out the same way in every market we work: Singapore, Malaysia, Australia, the US, and Canada. The operating-model logic does not change at a border. What changes is the exposure rate, which is a function of how early and how heavily AI answers reach a given market's buyers.

The English-first, high-search-maturity markets feel it first and hardest. In Singapore and Australia, buyers with strong AI-tool adoption are already arriving pre-shortlisted, having asked an assistant to narrow the field before they touch a brand's property. The US is the largest and most contested surface, so citation competition is fiercest there. Canada tracks the US closely. Malaysia's adoption curve runs a step behind in places, which buys a little more runway, not a different destination.

The wrong conclusion is to build five local playbooks. The right one is a single operating model, be the cited source and own the first-party relationship, applied with market-specific pacing. A Singapore team may need to move on citation-readiness this quarter while a Malaysia team has a quarter or two of runway to build the first-party spine first. Same model, different order of operations. That is how we run it: one system, five markets, not five systems.

How AI-exposed is your marketing model?

Before you rebuild anything, it helps to know how exposed your current model is to AI intermediation. This is a rough heuristic self-scorer, six questions, each weighted by how much it drives exposure. It returns a band and the gap to close first. It is a directional read, not a diagnosis, and it is not a promise of any outcome. The same six questions sit in the static table below for anyone who cannot run the widget.

AI-exposure self-scorer (heuristic)

Six questions. Answer honestly. The band is a rough heuristic signal, not a measurement or a guarantee.

1. What share of your discovery today comes from search and organic content? (higher share = more exposed)

2. Is your content built to be extracted and cited, or to rank and be clicked?

3. Does your measurement still rely mainly on last-click attribution?

4. How much of your audience relationship do you own directly (first-party data, owned channels)?

5. Do you track whether AI answer engines cite or mention your brand?

6. Are your buyers the kind who research with AI assistants before contacting you?

Static version of the self-scorer: the six exposure signals and what a high answer implies.
Signal A high-exposure answer looks like The fix it points to
Discovery mix Most discovery comes from search and organic content Rebuild top-intent answers to be citation-ready
Content design Content is built to rank and be clicked Shift to extractable, answer-shaped, evidenced chunks
Measurement basis Last-click is still the source of truth Add causal methods and first-party signal, read trends
Audience ownership Reach is mostly rented, owned channels are thin Stand up a first-party data spine and owned channels
Citation visibility No tracking of AI citations or mentions Start citation and brand-mention monitoring
Buyer behaviour Buyers research with AI assistants before contact Be the entity agents shortlist, not just a click destination

Where to start: the operator playbook

The failure mode we see most often is trying to rebuild everything at once, which stalls, and then concluding the whole thing was hype. It is not. It is a sequence. Diagnose, fix the widest gap, then compound. Here is the order we run it in, and it is deliberately conservative about how much changes at once.

  1. Diagnose your exposure, then rank the gaps

    Use the scorer above as a starting read, then get specific. Which of the six signals is your worst? That is where the money is, because the widest gap is where the repricing is doing the most damage while you are not looking. Rank them and pick one to move first. Do not spread thin across all six.

  2. Rebuild the answers to your ten highest-intent queries

    Not a content calendar. Ten questions your best buyers actually ask an assistant, rebuilt so your page is the one a model would quote: answer first, evidenced, structured, entity signals clear. This is the fastest lever, because a handful of citation-ready answers can start earning mentions in weeks. The GEO playbook is the how.

  3. Stand up a first-party data spine

    Decide what audience relationship you will own directly and start collecting it with consent. Logged-in behaviour, email engagement, self-reported source. This is the substrate that powers lifecycle and feeds cookie-independent measurement, and it is the asset no intermediary can take. Start small and real rather than waiting for the perfect stack.

  4. Move measurement off last-click before it lies

    Add one causal method and one citation-tracking habit. Not a full rebuild. A truth check, so you stop defunding work last-click cannot see.

  5. Tilt budget toward what compounds

    Treat paid reach as a repricing asset, not the growth engine. Move marginal budget toward owned channels, first-party data, and citation authority, because those compound while rented attention is skimmed off the top. Rebalance deliberately. Do not cut paid to zero and do not defend it out of habit.

  6. Follow the four pillars in the order your gaps demand

    The funnel, content, inbound, and lifecycle pillars each go deep on one rebuild. Read them in the order your diagnosis points to, not front to back. The team whose worst signal is measurement should not start with content.

None of this is a bet that AI killed marketing. It is the opposite claim. Digital marketing in the AI era rewards the disciplines it always rewarded, being genuinely useful, being trusted, owning the relationship, more than the old model ever did, because the intermediary in the middle is specifically built to reward sources it can trust and reuse. The teams that win are not the ones who spend the most. They are the ones the machine reaches for. If you want that rebuild run as a working system across your five markets rather than assembled from a stack of blog posts, that is the work we do. Start with how we work.

Frequently asked questions

How has digital marketing changed with AI?

The distribution layer moved. For twenty years the model was to win a rank, earn a click, and convert the visitor on your own site. AI answer engines now resolve a large share of queries inside the answer surface, so the query gets satisfied without the click. The job shifts from capturing the click to being the source the answer engine cites, and from renting reach to owning first-party signal. The classic disciplines still exist, but each one now runs through an AI intermediary that sits between you and the buyer.

What does AI-era marketing actually look like?

It looks like fewer, deeper assets built to be extracted and cited rather than to rank and be clicked, a measurement stack that tolerates zero-click influence instead of demanding a last-click trail, and a heavy tilt toward owned channels like email and community where you control the relationship. Paid still works, but you buy it knowing reach is repricing. The winners read as sources of record in their category: the site an AI assistant reaches for when a buyer asks it to compare, shortlist, or recommend.

Is digital marketing still worth it with AI search?

Yes, but the payoff curve changed shape. Cheap-reach tactics that leaned on volume of clicks are worth less, because the clicks are being intercepted. Work that builds durable entity authority and first-party signal is worth more, because it compounds and because AI systems reward sources they can trust and reuse. The mistake is treating this as a reason to spend less. It is a reason to spend differently: toward being cited and toward owning the relationship, away from renting attention that an answer engine now skims off the top.

What is zero-click and why does it matter for marketers?

Zero-click describes a query that gets answered inside the search or chat surface, so the user never clicks through to a source. AI Overviews and chat assistants generate an answer in place, often citing a handful of sources without sending most users onward. For marketers it matters because the click was the unit almost every funnel, attribution model, and content brief was built around. When the click disappears but your brand is still the cited source, you got the influence without the visit, and your measurement has to be rebuilt to see it. Our zero-click search strategy covers the tactics.

What is agentic shortlisting?

Agentic shortlisting is when an AI assistant does the buyer's early research for them: it reads across sources, assembles a shortlist of options, and hands back a recommendation with reasons. The buyer may never see the ten links a search page would have shown. They see the three the agent chose. If your brand is not in the model's working set of trusted, well-structured, citable sources, you are not on the shortlist, and no amount of retargeting reaches a buyer who never entered your funnel. Being the entity agents reach for is the new top-of-funnel.

Do I need to abandon SEO and paid ads in the AI era?

No. Search and paid are repricing, not vanishing. Traditional SEO still feeds the corpus that answer engines read from, so ranking well and being cited are correlated, not opposed. Paid still buys reach, and owned surfaces still need it as a feeder. What changes is the weighting and the goal. You optimise the same content to be extractable and citation-ready, you treat paid as a repricing asset rather than a growth engine you can lean on forever, and you move budget toward the owned and first-party work that compounds. Rebalance, do not abandon.

How do I measure marketing when clicks and cookies both decay?

Move up a level. Last-click attribution assumed a clean visit trail that zero-click and cookie decay both break. The durable stack pairs first-party signal you own (logged-in behaviour, email engagement, self-reported source) with aggregate causal methods like incrementality tests and marketing mix modelling that do not need user-level tracking. You also add citation and mention tracking to see the influence that never produced a click. The honest version accepts wider confidence ranges and reads trends over quarters instead of chasing a single deterministic path. See our first-party data strategy.

What is the single most important shift for a marketing team to make?

Stop optimising to be clicked and start optimising to be cited. Everything else follows from that one reframe. Content gets built in answer-shaped, extractable chunks with clear entity signals. Measurement stops demanding a click trail. Budget tilts toward owned channels and first-party data because those are the assets an answer engine cannot skim off the top. If a team makes only one change this quarter, it should be to pick its ten highest-intent queries and rebuild the answers to those so an AI assistant would quote your page over a competitor's.

Does this apply outside the US, in markets like Singapore or Australia?

Yes. AI Overviews and chat assistants are broadly available across Singapore, Malaysia, Australia, the US, and Canada, and the operating-model logic does not depend on a single market. The nuance is exposure rate: English-first, high-search-maturity markets tend to feel zero-click first and hardest, so a Singapore or Australia buyer may already be arriving pre-shortlisted by an assistant. The move is the same everywhere. Be the cited source in your category and own the first-party relationship. We work this across all five markets from one operating model rather than five local playbooks.

How is this different from just doing SEO or GEO?

Generative engine optimization (GEO) is the tactical craft of getting a page cited by answer engines. This pillar is the operating-model layer above it: how the whole marketing system, funnel, content, inbound, lifecycle, budget, and measurement, gets repriced and rebuilt, with GEO as one instrument inside it. Think of GEO as how you win a citation and this as why citations and first-party signal replace clicks and reach as the assets you build the business on. Our GEO playbook is the tactical companion to this strategic frame.

Where should a team actually start?

Start with a diagnosis, not a rebuild. Score how exposed your current model is to AI intermediation, then fix the widest gap first. In practice that is usually one of three: rebuild the answers to your top-intent queries so they are citation-ready, stand up a first-party data spine so you are not renting your audience, or move measurement off last-click before it lies to you. Pick the one where the gap is widest and the fix is fastest. The self-scorer in this piece is a rough starting read, not a plan.

Related

Work with leapbuzz

The self-scorer flagged your exposure. Want the rebuild that comes after the diagnosis?

leapbuzz rebuilds marketing operating models for the AI era across Singapore, Malaysia, Australia, the US, and Canada. Citation-ready content, a first-party data spine, and measurement that survives zero-click, built as a working system rather than a slide deck.

Talk to us