Growth Hacking Is Still Real in 2026: The Evidence, Including What Argues Against It
Is growth hacking dead in 2026? The honest answer, with sourced data on rising CAC, the 10x retention spread between operators, and the strongest case against.
the short answer: growth hacking is not dead, but the reason it survives is the opposite of the reason people usually give. it is not that “the term evolved into growth marketing.” it is that the gap between operators who run structured experiments and operators who don’t got measurably wider in 2025 and 2026. when outcomes spread out under identical market conditions, that spread is the evidence the practice matters.
here is the case, the numbers behind it, and the strongest argument against it, which i think is partly right.
What actually got harder
three things changed, and all three are real.
search stopped sending clicks. 68.01% of google searches ended without a click between january and april 2026, up from 60.45% in 2024 (sparktoro and similarweb clickstream panel, published 9 june 2026). pew research, looking at 68,000 real queries, found users clicked a result 8% of the time when an AI summary appeared versus 15% when it didn’t. worth noting: google disputes pew’s methodology, arguing the study window overlapped unrelated algorithm testing. i think the direction is right even if the magnitude is arguable.
acquisition got expensive. SaaS customer acquisition cost is up 40 to 60% since 2023. the median SaaS company now spends $2.00 to acquire $1 of new annual recurring revenue, up 14% from 2023. on mobile, average cost per install has risen 15 to 25% year over year since 2023.
the content channel flooded. ahrefs studied 900,000 newly created pages and found 74% contain some AI-generated content. cloudflare reported bots generated 57.5% of web requests in 2026, more than humans.
if you stopped reading there you would conclude growth work is finished. that reading misses the more interesting number.
The number that actually matters
in the same period that SaaS CAC rose 40 to 60%, companies that adopted AI across their acquisition stack reported CAC reductions of 30 to 47%.
same market. same year. opposite outcomes.
two more spreads point the same way:
- brands testing 10 or more creative concepts per month achieve 31% lower cost per acquisition than brands testing fewer than five. that comes from an analysis of over 500,000 meta ads representing more than $1B in spend across 6,000 brands.
- among 3,519 consumer AI apps covering more than 50 million paid subscriptions, high-retention apps keep 13.9% of paying subscribers at twelve months. low-retention apps keep 1.4%. that is a tenfold spread, and revenuecat found the fork happens at the first renewal: 57.9% of monthly subscribers renew at the high-retention apps versus 30.2% at the low ones.
a tenfold difference in outcome under the same conditions is not luck and it is not budget. it is whether somebody ran the experiments.
What this looked like in practice
i scaled ZuAI from 10K to 2M users in 11 months at a blended CAC of $0.02, managing around $300k a month in ad spend and testing 150+ creatives a month.
that creative number is the part worth checking, because it is the mechanism rather than the result. the published benchmark for meta creative production is roughly one new ad per $3,000 of monthly spend. at $300k a month that predicts about 100 ads a month. we ran 150+. only about 5% of meta creatives become winners in 2026, so three winners a month requires testing 20 to 25 concepts minimum, and we wanted more than three.
the CAC was not a clever trick. it was throughput plus a rule that paid budget only followed creatives that had already proven themselves organically. the full breakdown is in the ZuAI case study.
The strongest argument against me
rand fishkin has made this case for years and he is not easy to dismiss.
his objection is empirical: businesses can spend four years perfecting a growth hack that only works for four months. hacks work briefly, then sputter, and the effort compounds into nothing. his alternative is the marketing flywheel, a mechanism where each effort makes the next one spin faster, which he argues is almost always better than a hack. he also rejects the framing itself, calling growth hacking silicon valley’s attempt to masculinize the concept of marketing.
three more objections hold up:
the canon is unrepeatable. every growth hacking article still leads with dropbox’s referral program, airbnb scraping craigslist, and hotmail’s email signature. those are from 2008 to 2010. airbnb’s was scraping a competitor’s platform. dropbox’s ran when storage was scarce and acquisition was nearly free. that the field still cites 15-year-old examples is itself evidence about how many comparable wins it has produced since.
the founder of the discipline has moved. sean ellis, who coined the term in 2010, said in a january 2025 interview that organic word of mouth and retention are better growth indicators than surface tactics like referral programs. the single most-cited growth hack of all time, demoted by the person who named the field.
specific tactics genuinely died. high-volume cold email stopped working when google, yahoo and microsoft moved sender requirements from recommended to enforced on 5 may 2025. product hunt now grants featured status to roughly 10% of launches, down from 60 to 98% in 2020 to 2023, and featured status determines most of the outcome. app store keyword stuffing no longer produces ranking lift. reddit spray-and-pray gets caught by systems that block 23 million spam views a day.
Where I think fishkin is right, and where I disagree
he is right that hacks are a bad unit of investment. he is right that flywheels beat tricks. i would not argue either point.
where i part company is the conclusion. the 10x retention spread and the 31% CPA gap did not appear because some companies found tricks. they appeared because some companies ran disciplined experiments and most did not.
experiments are how you find the flywheel. they are not a substitute for having one.
that reconciliation is the whole thing. if you run experiments hoping one becomes a permanent channel, fishkin is describing your future accurately. if you run experiments to find the loop your product actually supports, and then invest in the loop, the experiments were the cheapest research you will ever buy.
What changed about the work itself
the mechanism did not change. run a structured experiment, kill the losers, compound the winners. what changed is the half life.
platform gaps close in months now, not years. that makes experiment throughput the asset rather than any individual experiment, which is a stronger argument for the discipline rather than a weaker one. the operator who can run 20 tests a quarter beats the operator who found one good channel in 2024 and is still riding it.
and new surfaces keep opening. openai began accepting third-party app submissions for ChatGPT in early 2026, putting a distribution surface in front of roughly 800 million weekly users, larger than the apple app store’s installed base. apple is adding a second search-results ad placement in march 2026. every one of those is a cold start land grab, which is exactly the condition that produced the canonical wins everyone still quotes.
So is it real
yes, and the honest version is less flattering than the marketing version.
growth hacking in 2026 is not a set of clever tricks. it is the willingness to test at a volume most teams find tedious, kill things you liked, and write down the pass or fail line before you spend the money. the operators doing that are pulling away from the ones who aren’t, and the gap is measurable.
if you want that applied to your product rather than described at you, let’s talk. the breakdown call is free and you keep the plan either way.
sources: sparktoro and similarweb zero-click study (june 2026) · pew research center AI overview click-through study (68,000 queries) · ahrefs AI content study (900,000 pages) · revenuecat AI app retention study (3,519 apps, 50M+ subscriptions) · adliftr meta creative testing analysis (500,000+ ads, $1B+ spend) · sean ellis interview, january 2025 · rand fishkin on growth hacking and marketing flywheels.