Growth Hacking vs Growth Marketing in the AI Era

We might be over the Growth Hacking vs Growth Marketing debate by 2026, but proper, sustainable Growth has a new enemy now: vibe marketing.

A two-person team at a YC startup spins up a landing page in an afternoon, an entire newsletter sequence overnight, and forty ad variations before lunch. None of it is written by a human. None of it is tested against real product data. None of it is ever going to be opened by anyone, because the audience is also drowning in the same machine-made stuff.

Growth marketing in 2026 isn’t short of activity. But it’s often short of signal – making it tone-deaf and irrelevant.

So while the growth hacking vs. growth marketing debate had been resolved it seems (in growth marketing’s favour) – we need to address how “vibe coding” and “vibe marketing” is the new growth marketing – and what we should do about it to prevent you from derailing your growth strategy.

So this article:

• Makes the case that vibe marketing is just growth hacking with an AI wrapper, and is going to age the same way.
• Walks through five ways to use AI for growth marketing that actually compound, drawn from how teams like Lovable have grown and how MLforSEO’s Lazarina Stoy thinks about responsible AI use.
• Explains how MCP is changing the math for growth teams, including what we’re shipping at Userpilot to make our product data accessible to whatever AI agent you’re already using.

Growth marketing in 2026: the short version

For those of you who don’t have time to read the whole thing, or your agents 😅, here’s the summary:

The original debate

• Growth hacking = short-term, acquisition-focused, often manipulative tactics aimed at fast user growth.
• Growth marketing = full-funnel, retention-and-loyalty-driven, compounds slowly over the customer lifecycle.
• By 2024, growth hacking effectively lost its halo, and “growth marketing” became the default term.

The new chapter: vibe marketing

• Coined in early 2025, vibe marketing means describing a campaign in natural language and letting AI agents handle the execution: copy, visuals, landing pages, ad variants, automations.
• It collapses production cycles from weeks to days and lets two-person teams ship like ten-person teams.
• It’s also reproducing every mistake growth hacking ever made, just faster: optimising for volume over signal, ignoring the audience, treating shipping as the goal.

The responsible alternative: AI on real product data

• Use AI for the work that compounds: behavior-triggered messaging, predictive churn nudges, in-product personalisation, agentic experimentation, content built from real customer language.
• Connect AI to your product, not just your content stack. The MCP (Model Context Protocol) standard now lets any agent pull live product usage, session replay, NPS and survey data from your existing tools, including Userpilot.
• The shift Yazan Sehwail (our CEO) described in our recent interview is the right one: “You’re no longer operating. The AI is operating. You’re just basically evaluating and monitoring the agent workflow.”

The growth hacking vs. growth marketing debate, briefly

Quick refresher, because we’ll spend the rest of the post moving past this. If you already know the difference, skip ahead.

Growth hacking was a term coined by Sean Ellis around 2010 to describe the kind of work he did at Dropbox: scrappy, small-team, acquisition-focused experiments designed to spike user growth fast. It’s where the famous moves came from. Hotmail putting “Get your free email at Hotmail” in every outgoing email signature. Dropbox giving you extra storage for inviting friends. Airbnb’s reverse-engineered Craigslist integration.

Growth marketing is the broader, slower discipline that grew up in response. It pays attention to the whole journey from acquisition through activation, retention, expansion, and referral, not just the top of the funnel. It treats the product itself as a marketing channel. It uses behavioural data to personalise the experience inside the product, and it treats every stage of the lifecycle as a place to compound value.

The standard list of differences is well-known:

• Time horizon: growth hacking thinks in weeks; growth marketing thinks in quarters and years.
• Funnel coverage: growth hacking lives at the top of the funnel; growth marketing lives across the whole journey.
• Metric set: growth hacking tracks signups and virality; growth marketing tracks LTV, retention, expansion, and NPS.
• Channel: growth hacking is mostly external (web, paid, social); growth marketing is increasingly in-product, using in-app messaging and onboarding to drive outcomes.
• Risk profile: growth hacking optimises for variance; growth marketing optimises for compounding.

The honest version of this distinction is that they’re not opposites. Growth hacking is a subset of tactics; growth marketing is the strategy that decides which of those tactics are worth running. The interesting question was never “which one is right for you”, it was “when did the industry stop tolerating the worst of growth hacking, and why.”

Why growth hacking lost (and what got dropped from the conversation)

Growth marketing won the debate not because growth marketing arguments were better, but because growth hacking ran out of road. Three things happened in roughly 2020 to 2024.

Regulators caught up. Dark patterns (pre-ticked checkboxes, fake countdown timers, hidden cancellation flows) got specifically named and made finable in California (CCPA), India (DPDPA), the EU (Digital Services Act), and most major commerce jurisdictions. The “we just A/B tested into manipulation” defence stopped working in court. A 2024 ACM survey documented just how much of the early growth hacking playbook had quietly become illegal.

The metrics turned. Once teams started measuring D90 retention and 12-month LTV instead of just signup count, the dark-patterned funnels stopped looking impressive. A short-term acquisition bump that costs you a 30% drop in retention isn’t a growth hack; it’s a slow churn machine with a marketing budget.

Trust got priced. Elena Verna, now Head of Growth at Lovable, has been making a version of this argument for years and crystallised it in her January 2026 essay “Growth Is Now a Trust Problem”: distribution channels are saturated, “marketing tactics” are losing leverage, and the only durable advantage left is being a brand people actually trust to do what it says it does. In an environment where the audience is jaded and the channels are crowded, the growth hacker’s bag of tricks just doesn’t pay anymore.

By the end of 2024, “growth marketing” had quietly become the default label for the work, and “growth hacker” had become a slightly cringe-y LinkedIn relic. The debate had been won.

And then it came back.

Vibe marketing is just growth hacking with an AI wrapper

In February 2025, James Dickerson was sharing AI-marketing automations in a private Slack community. Greg Isenberg saw what he was building and told him to call it “vibe marketing,” riffing on Andrej Karpathy’s “vibe coding.” Within eight months, search interest in “vibe marketing” was up 686%, Fortune 500s were running internal workshops, and Ramp had posted a Vibe Growth Marketing Manager role.

The official definition is something like: vibe marketing is using AI to set a brand’s tone, automate content creation and testing, and ship campaigns faster and cheaper. You describe what you want in plain language, the AI agent handles the execution: copy, design, landing pages, ad variations, automations. The pitch is that production cycles collapse from 4 to 6 weeks down to 3 to 6 days, that two-person teams can do the work of ten, and that “taste” replaces “specialist skill” as the bottleneck.

I want to be precise here: some of this is real and good. Multimodal content production has genuinely become an order of magnitude cheaper. Smart prompting and a good system-prompt design can compress weeks of testing into days. Tiny teams can punch above their weight. None of that is going away.

But the structural shape of vibe marketing (fast, individual, output-maximising, optimising for production speed rather than audience signal) is identical to the structural shape of the growth hacking that the industry just spent five years walking away from.

It is going to age the same way, for three reasons.

1. Output is being decoupled from signal, again. The whole logic of vibe marketing is “ship more, faster.” But the audience hasn’t gotten any larger, and the inbox hasn’t gotten any longer. Heinz Marketing’s analysis of the B2B “AI slop” backlash argues that nearly 90% of audiences now say they prefer human-made content, that “AI booing” is a real and rising phenomenon, and that the future of marketing isn’t who can generate the most content, but who can apply the strongest human filter on top of AI. As Yazan Sehwail, our CEO, put it about the parallel problem in product: “As producing and building features become a lot cheaper, instead of every quarter you’re releasing one or two, now you’re releasing 7, 8, 9. It becomes even harder to manually track each one.” Same thing happens to marketing teams shipping 50 to 100 pieces of content a month: the team can ship it, but nobody can tell what’s working.

2. The dark patterns are coming back, just dressed up. Auto-personalised landing pages that use scraped LinkedIn data without consent, “AI SDRs” that mimic human follow-up, generative-AI ads that fabricate testimonials or use deepfaked customer faces, churn-prevention agents that hide cancellation behind a chat loop. None of this looks like a 2015 growth hack on the surface. All of it has the same structure underneath. And the same regulators that finally caught up with dark patterns are moving on AI-specific manipulation now (the EU AI Act, FTC’s “Operation AI Comply,” KPMG’s tracking of consumer trust in AI-generated content sitting at ~50%).

3. There is a real environmental cost the deck doesn’t show. This is the part the vibe marketing tutorials never discuss, and it’s where the growth hacking analogy actually breaks down. Vibe marketing is worse, because vibe marketing has a sustainability bill that growth hacking never did.

The sustainability case for responsible AI in growth marketing

The vibe marketing pitch deck talks a lot about velocity and almost never about cost. But the cost is real, it’s measurable, and it’s hitting the marketing industry harder than most others.

The clearest articulation I’ve seen of this argument comes from Lazarina Stoy, an SEO and ML practitioner who runs MLforSEO. Her BrightonSEO April 2026 talk, “The sustainability issue with AI and what we can change as marketers,” lays out the receipts.

The unit costs of AI work are higher than most marketers realise:

 

• One generated image uses roughly the energy needed to fully charge a smartphone.
• One ChatGPT query uses about 10× the energy of a Google search (2.9 Wh vs. 0.3 Wh, IEA 2025) and roughly 250 to 500 ml of water for server cooling per 100-word output.
• ChatGPT processes ~2.5 billion prompts per day as of 2026 (DemandSage / Superlines / OpenAI).
• Projected data center electricity consumption hits ~1,050 TWh by 2026 (IEA, MIT, Lawrence Berkeley National Lab), more than Japan’s entire national consumption.

The marketing industry compounds this in three specific ways:

Query fan-out. When a human searches for something, they visit maybe five websites. When an AI agent fulfils the same request, Cloudflare’s CEO told SXSW it visits ~5,000: a roughly 1,000× multiplication factor per query. AI agent traffic was up 7,851% year-over-year by late 2025. AI bots alone now hit Cloudflare’s network with ~50 billion crawler requests per day.

AI Overviews and re-prompting. AI Overviews now appear in ~25% of Google search results. Each one is a downstream LLM call. Multiply 10× the energy of a regular search × ~8.5 billion daily searches × 1 to 3 follow-up re-prompts and you get, in Lazarina’s words, “a massive, invisible environmental impact.”

The “everyone is a marketer now” problem. The same democratisation that made vibe coding and vibe marketing possible also turned every individual marketer into a potential model-trainer, custom-tool-builder, and one-off-script generator. As Lazarina puts it, products and tools and websites and projects “are being built every day, because they can be built, not because they should exist.” It’s the AI version of the feature factory, and it’s a feature factory with an electricity bill.

The downstream effects on trust are now showing up in the data. Ahrefs analysed 900,000 newly-published pages in April 2025 and found AI-generated content in 74% of them. KPMG’s 2025 trust survey put consumer distrust of AI-generated responses at 50%. Reddit co-founder Alexis Ohanian has said openly: “the dead internet theory is real.” MIT’s research on AI reliance describes “cognitive atrophy” in heavy users; Harvard’s recent work warns that “if AI does your thinking, it undercuts creativity.”

Lazarina’s prescription is the cleanest version of the responsible-AI framework I’ve seen, and it maps directly onto growth marketing: be AI-assisted, not AI-driven.

Five ways to use AI for growth marketing without becoming the new growth hacker

The pattern across all five of these is the same: AI is most useful when it’s pointed at real product behaviour from real users, not at a blank page where it has to invent something to say. The work that compounds is the work that’s grounded in what your customers are actually doing inside your product.

1. Behaviour-triggered messaging that responds to what users actually did

The unsustainable version: blast the entire list with the same AI-written email. The compounding version: trigger a small, specific message when a user does something specific in the product. A user who hits a usage limit gets one message. A user who just discovered a feature gets a different one. A user who hasn’t logged in for two weeks but used to be a power user gets a third.

This is where Loom’s contextual upgrade prompt is doing real work. It shows up at the moment the user is most likely to want the paid feature, not in a Tuesday-morning newsletter.

2. Predictive churn and expansion signals (and acting on them automatically)

AI is genuinely good at spotting patterns in usage data that humans can’t see at scale. A drop in invoice usage. A product manager who’s stopped opening reports. An admin who hasn’t added a new seat in 90 days. The use case isn’t “show me a chart of churn risk.” It’s “trigger an in-app message, an account note, or a CSM task automatically when the model says risk just jumped.”

3. In-product personalisation built from real onboarding signals

The growth hacking version of personalisation was “first name in subject line.” The growth marketing version was branched onboarding that asked who you were. The AI version is to take both (declared role, observed behaviour, account stage) and let an agent assemble the in-app experience around what the user is actually trying to do.

Notion’s onboarding survey is still one of the cleanest examples in the wild: it asks two or three questions, then routes you into different first-time experiences based on the answers. Now imagine that same logic, but the experience itself is being assembled by an agent in response to your actual usage signal, not from a static decision tree.

💡 Read related: How to build a personalised customer experience that actually moves retention.

4. Agentic A/B testing that runs continuously, not in two-arm sprints

Classic A/B testing assumes you have time to set up two variants, wait three weeks, declare a winner, and ship. Vibe marketing turned that into “test 40 ad variations simultaneously and let the algorithm pick.” Useful for paid social, mostly useless for in-product changes that you actually have to maintain.

The compounding version of this is to use an agent to continuously A/B test in-product flows on a tight, well-instrumented surface (modals, onboarding checklists, upgrade prompts) where you can measure the outcome that matters, not just the click. As Abrar Abutouq, one of our PMs, put it when I asked her about AI suggestions: “Sometimes [AI] might be right, might be wrong, but I always believe that every single suggestion you gather from feedback, from AI or a teammate, it’s an experiment. You need to try it out. The product is all about experiments.”

5. Content built from real customer language (not blank-page generation)

Most AI content generation today is what Lazarina would call “blank-page prompting”: tell the model the topic, hope for the best. The output is fluent and forgettable, which is exactly the slop problem.

The version that compounds is using AI to extract language from places your customers are already using your product (support tickets, in-app surveys, sales calls, session replays) and turning that into content that uses their actual words, their actual objections, their actual outcomes.

Lovable’s growth team has been very public that this is how they write: their “Zero to $10M ARR in 2 months” post reads less like marketing copy and more like a transcript of what their early users were saying out loud. Elena Verna’s recent Lenny’s Podcast appearance on Lovable’s playbook reinforces the point: build-in-public, customer-language-driven content out-distributes anything generated from a content-calendar prompt.

How MCP changes the math, and what we’re doing 

The single biggest reason 2026 isn’t just a re-run of 2015 is the Model Context Protocol. MCP is an open standard, originally proposed by Anthropic in late 2024 and now adopted by basically everyone (OpenAI, Google, Microsoft, the major analytics and CRM vendors).

It does one thing: it lets any AI agent securely pull live data from any tool that ships an MCP server, without bespoke integrations.

For growth marketers, that’s the difference between “AI that writes about your product” and “AI that knows what your users are doing in your product right now.”

Once your product data is sitting behind an MCP server, the AI agent can do three things growth marketers couldn’t do before:

• Respond to user behaviour in real time. A Workflow in Userpilot listens for the behavioural signal, the agent decides the right response, and the message ships in the right channel (in-app, email, mobile push) without a human in the loop on the easy cases.
• Send timely, situationally aware emails. Not “blast the segment on Tuesday.” More like: “this user just hit usage limit X for the third time this week, send them the upgrade context they actually need.”
• Generate bespoke personalised content for each user, grounded in their actual data. The same modal copy, written with the user’s own usage patterns and account context, instead of from a blank page.

This is the part that maps directly to our CEO’s, Yazan Sehwail’s vision for where AI in B2B SaaS is going. He put it like this in our recent interview:

“If you as a marketer wanted to see, using session replay, NPS data, survey data, and product usage data, you’re able to get your answer without having to go to Userpilot, without having to pull data and upload it to someone. So this is why MCP is gonna be a game changer. As teams start deploying their own AI agents, those agents are gonna tap on our existing infrastructure that will be powering all of the usage and all the product data, and that’s extremely powerful.”

 

What this looks like in practice 

Lia, our AI agent, does the work end-to-end. When we first started building Lia, Our CEO’s initial design was an AI assistant: you’d ask it to create a report, then ask it to analyse the report, then ask it to suggest a flow. He scrapped that model entirely in December 2024. His exact words:

“People don’t wanna do any of this. That’s the truth. What it’s gonna be is that you literally do not need to do anything. It’s gonna look like you just go, you create a project, you tell it what you want, and it should do the rest. You’re no longer operating. The AI is operating. You’re just basically evaluating and monitoring the agent workflow.”

That’s the right shape for AI in growth marketing. You set the goal (“improve trial-to-paid conversion for the EMEA cohort”), and the agent pulls the data, builds the analysis, surfaces the friction points, and proposes the in-app experiences that would address them. You evaluate; you don’t operate.

Workflows make the responses real. Once Lia has surfaced a recommendation, Workflows is what actually ships the response: the in-app modal, the triggered email, the mobile push.

The growth marketer’s job stops being “manually build flows for every cohort” and starts being “approve the agent’s draft.”

Concrete example from inside the building. Abrar Abutouq, one of our PMs, recently shipped a clean version of this loop for our email feature. She noticed in the funnel that users were dropping off at the domain-verification step, the gating step needed to unlock the feature. Instead of queuing an engineering ticket, she did this:

“Within a few hours, I just created a targeting tooltip and showed it to users and highlighted the correct steps for them to make it clear what to do next. That helped a lot on reducing friction and supporting users in real time without involving our dev team.”

Abrar Abutouq, Product Manager, Userpilot

That’s the loop in miniature: real product data spotted the friction, a no-code in-app fix shipped within hours, the funnel improved, no engineer was paged, no AI hallucinated anything. It’s a deliberately small example. Most growth marketing wins are deliberately small examples that compound.

💡 Read related: PLG in the AI era, what’s actually changed.

Vibe marketing is the new growth hacking. AI on real product data is the new growth marketing.

Growth hacking earned its bad reputation honestly. The good ideas it contributed (referrals, in-product virality, low-cost acquisition experiments) are still in everyone’s playbook. The bad ones (dark patterns, manipulation, output-without-signal) got regulated, metric’d, and culturally walked away from.

Vibe marketing is on the same arc, on a faster clock. The good ideas (multimodal production, system-prompted workflows, two-person teams shipping like ten-person teams) are real and worth keeping. The bad ones (output-without-signal, audience-as-broadcast-target, the genuine and growing environmental cost) are going to age the same way. The regulators are already moving. The audience already knows. The data center bill is already being written.

The version that’s worth investing in is the one Lazarina described in her BrightonSEO talk and the one Yazan described in our interview, from two completely different angles: AI-assisted, not AI-driven.

Humans set the direction; agents do the work; the data is real product data, not a blank page. Connect AI to your product via MCP. Trigger it on real user behaviour. Use it to act, not to broadcast.

That’s what growth marketing looks like in 2026. It’s quieter than vibe marketing. It compounds harder. It doesn’t end up in the next “AI booing” wave. And it’s the version of the playbook that’s still going to be useful in 2028.

If you want to see how this looks inside one platform (Lia, MCP Server, Workflows, and the underlying analytics + session replay + survey data that grounds all of it), you can book a Userpilot demo. We’d rather show you the loop than tell you about it.