What a customer actually costs: CAC by channel for a one-person AI product

A new analysis of customer acquisition cost (CAC) by channel for one-person AI products finds that paid channels run hundreds of dollars per customer, with B2B paid search near $800 and LinkedIn near $982, while earned channels like SEO cost around $205 but face erosion from AI Overviews that have taken 58% of clicks from the top result. The analysis, based on data from sources including FirstPageSage, Ahrefs, RevenueCat, and David Skok's SaaS Metrics, warns that AI products churn about 30% faster than non-AI apps, with 6.1% twelve-month retention vs. 9.5%, making it harder to recover CAC before customers leave.

The Study · Comparison What a customer actually costs: CAC by channel for a one-person AI product Search “customer acquisition cost by channel” and you get a tidy table: referral cheapest, paid search in the middle, LinkedIn at the top. Look at who published it and the table is almost always an agency that sells one of those channels. The most-cited CAC-by-channel numbers come from a company that sells SEO and content. The referral numbers come from companies that sell referral software. The numbers are not fake, but they are marketing, and they are aimed at getting you to pick the channel the publisher happens to rent. The answer, for a one-person AI product, is the channel your own time can run. So this is one table built from the least self-interested sources I could find, with every vendor figure flagged as directional, judged against one question a one-person AI product actually has to answer: can you recover the cost before the customer churns? Because the catch that none of the channel tables price is that an AI product is buying revenue that leaks faster than the SaaS the benchmarks were built on. The short version If you are choosing where to spend, the numbers land like this: Paid runs hundreds of dollars a customer. B2B paid search sits near $800 https://firstpagesage.com , LinkedIn near $982 with a $4-5k monthly floor just to get a signal, and Meta’s cheap clicks are mostly unmeasurable since Apple’s tracking prompt. Earned is cheaper, slower, and eroding. SEO lands around $205 https://firstpagesage.com with a seven-month payback, while AI Overviews have taken 58% of the clicks https://ahrefs.com/blog/ai-overviews-reduce-clicks-update/ off the top result. The AI catch shrinks every payback. Consumer AI churns about 30% faster https://www.revenuecat.com/state-of-subscription-apps-2026/ than non-AI apps, worst under $50 a month, so the same CAC repays from a thinner stream. Founder hours are a cost the tables skip. Content CAC swings between about $20 and $180 https://www.growandconvert.com/content-marketing/cost-of-content-marketing/ depending on whose salary does the writing; a free channel bills you in time. The order of operations: owned email first, community and replies as feeders, referral once users retain, and paid only after LTV clears the 3:1 bar https://www.forentrepreneurs.com/saas-metrics-2/ . The channel-by-channel evidence is below. The number is always for sale Two anchors hold the whole piece up, and neither is selling a channel. The first is the viability bar. The rule that a customer should be worth about three times what they cost to acquire, and that you should recover the cost in under a year, traces to David Skok’s SaaS Metrics https://www.forentrepreneurs.com/saas-metrics-2/ , written by an investor describing what makes a business survivable, which beats an agency describing what to buy. Treat 3:1 LTV to CAC and sub-twelve-month payback as a floor. The best SaaS recovers CAC in five to seven months. The second is the AI catch, and it sits on the LTV side of that ratio. Consumer AI apps churn about 30 percent faster https://www.revenuecat.com/state-of-subscription-apps-2026/ than non-AI apps, per RevenueCat’s 2026 report across more than 115,000 apps: 6.1 percent twelve-month retention against 9.5 percent. The cheap end is worse still. AI products under $50 a month show gross revenue retention around 23 percent https://www.growthunhinged.com , the weakest of any tier, while $250-plus AI products retain like normal SaaS. Add gross margins that run roughly 45 to 52 percent https://www.saastr.com against SaaS’s 80 to 90, because the model provider takes a cut of every call, and the same CAC has to be repaid from a thinner, faster-draining stream. Every channel below inherits that problem. Here is the table. Read the catch column first. | Channel | Directional CAC | The catch | |---|---|---| | Paid search Google | ~$800 B2B SaaS | Real buyer intent, but only clears above meaningful LTV | | Paid social Meta / X | ~$230; cheapest clicks | Cheap clicks are not ROI, and you cannot measure the CAC | | LinkedIn ads | ~$982; $100-150 per lead | Needs $4-5k/mo just for a signal | | SEO / content | ~$205; ~7-month payback | Front-loaded, slow, and the clicks are eroding | | Owned newsletter | near-zero cash | The list is the asset, but building it is time | | Founder-led | near-zero cash | Reach is mostly other founders, rarely buyers | | Community / forum | ~$0 cash | Rules-fragile, time-heavy, a feeder not a sale | | Referral | ~$23 cheaper, +16-25% LTV | A multiplier that needs retained customers first | The dollar CACs for paid search, LinkedIn, and referral come from First Page Sage https://firstpagesage.com , an agency that sells organic channels, built from multi-year client averages that fold in a learning period. Read them as order-of-magnitude. The paid channels, sorted by buyer intent Rank paid by buyer intent and the order flips from what the CPC suggests. Google Search sits in the middle on cost and top on intent, because someone typing your category is looking to buy. WordStream’s 2025 benchmarks https://www.wordstream.com/blog/ws/2022/05/18/search-advertising-benchmarks put the all-industry cost per lead near $70, and business services above $100. But a lead is not a customer, and software landing pages convert visitors to signups at a median around 3.8 percent https://unbounce.com/conversion-benchmark-report/ per Unbounce, below the all-industry 6.6. Multiply it out and the real customer CAC lands in the hundreds. Worth it when your LTV is real. Underwater when you are charging $19 a month. Meta and X are the cheapest raw traffic, and the biggest trap. Meta traffic clicks run around $0.70 https://www.wordstream.com/facebook-advertising-benchmarks and lead-form leads under $30. X’s CPM sits at or below Meta’s now, with clicks well under a dollar. Cheap. The problem is that you cannot tell what any of it did, because after Apple’s tracking prompt about 86 percent of users opt out, so channel-level attribution mostly does not exist at a solo operator’s volume. You will spend $400, gain some signups, and never know which came from the ad. Cheap clicks you cannot measure are a way to lose money slowly with a dashboard that looks busy. LinkedIn is the most expensive lead in software, and sometimes worth it, almost never for you. HockeyStack’s 2025 benchmark https://www.hockeystack.com across tens of millions in SaaS spend puts clicks above $8 and cost per lead in the $100-to-$150 range, and the practical floor to get a readable signal is four to five thousand dollars a month. That math works above a few-thousand-dollar contract value. It prices out nearly every one-person AI product. The verdict on paid: it is a tool for scaling a channel that already converts, and a bad way to find out whether anyone wants the thing. At seed stage, paid mostly buys you an unmeasurable number and a smaller bank balance. The earned channels, and why cheap has a time bill Earned channels are cheaper in cash and expensive in the one thing a solo founder cannot buy back. SEO and content beat paid, but by less than the brochures claim. The credible gap for B2B software is about $205 organic against $341 paid https://firstpagesage.com , a 20-to-50 percent saving, far short of the five-to-ten-times the category pages imply, and a content program takes around seven months to break even. And the ground is moving under it. Google’s AI Overviews cut clicks to the top organic result by 58 percent https://ahrefs.com/blog/ai-overviews-reduce-clicks-update/ as of early 2026 per Ahrefs, with position-one click-through falling from 7.6 to 1.6 percent. Independent Pew data https://www.pewresearch.org shows users click a normal result about half as often when an AI summary appears. Close to 58 percent of US searches already end with no click at all, and roughly 83 percent when an AI Overview is present. The work is shifting from earning the click to being the source the AI cites /study/what-a-chatgpt-citation-is-worth/ , which is a different skill and a moving target. Owned email is the workhorse of the whole table. A list converts to paid at something like 2 to 5 percent https://www.activecampaign.com , against a fraction of a percent for raw social reach, so a small list can out-earn a large following. Bought subscribers run a dollar or two each, which means the cash-cheap path is building the list yourself over time. That time is the cost, and it is real, but the list is the one asset every other channel on this table should be feeding. Founder-led building gets one row here on purpose, because it has its own piece. Raw followers convert to customers at a fraction of a percent, the audience skews toward other builders rather than buyers, and a useful reply on someone’s frustration thread converts far better than a broadcast post. The full funnel math is in the build-in-public study /study/build-in-public-revenue/ ; on this table it is a near-zero-cash, high-time feeder. Community and referral round out the cheap end. There are real $0-spend community wins, one documented case reached about $17,000 in monthly revenue in four months on value-first posting alone, though that is one self-reported story and no benchmark, and community marketing is fragile: a promotional post from a new account gets removed or shadowbanned fast. MicroConf’s survey, via Freemius https://freemius.com , found 47 percent of indie founders https://freemius.com name integrations and communities as their most dependable growth source, having given up on unmeasurable paid ads. Referral is the strongest cheap channel when it fires: the primary Wharton study https://www.journals.sagepub.com of about 10,000 customers found referred customers cost roughly $23 less to acquire and were worth 16 to 25 percent more over time. That is 2011 banking data, so treat the magnitudes as directional for a 2026 AI product, but the direction is well established. The catch is that referral is a multiplier on happy, retained customers, and a fast-churning product may never accumulate enough of them to turn it on. The bar every channel has to clear Lay the viability bar over the table and most paid options fall off for a low-priced AI product. The bar is 3:1 LTV to CAC and payback under twelve months, five to seven for the best SaaS. A self-serve product should recover CAC in one to seven months. So a paid channel that cannot pay itself back inside a year is a leak. Now stack the AI catch on top. The revenue you are buying churns about 30 percent faster than the SaaS these benchmarks assume, the sub-$50 tier retains worst of all, and the margin on each dollar is roughly half. The same $230 Meta CAC that a healthy SaaS repays comfortably has to come back out of a customer who is likelier to cancel this quarter and whose payment is half margin /study/unit-economics-of-wrapping-an-llm/ . The generic benchmark says the channel works. The AI-adjusted math says it might not. This is why the cheapest click can still be the worst buy: what you acquired may cancel before it ever pays you back. Your time is the CAC nobody prices Here is the number that reframes the whole table. Grow & Convert modeled the CAC of content marketing and found it swings from about $20 to about $180 https://www.growandconvert.com/content-marketing/cost-of-content-marketing/ per customer purely by changing what you pay the person writing, from nothing to $10,000 a month. The channel did not change. The founder’s salary did. A “free” channel is an opportunity-cost subsidy you are paying out of your own hours, and for a solo founder those hours are the scarcest input there is. That is the clear-eyed way to read the cheap end of the table, and it cuts both directions. Founder-led and community cost no cash and a great deal of time, so they are cheap only if your time is worth little, and expensive the moment it is worth a lot. Paid costs cash and little time, which is exactly why it is the right tool once a channel converts and you want to stop being the channel. Two disciplines keep this from fooling you. First, blended CAC lies. Hustle Fund’s worked example https://www.hustlefund.vc shows a true $100 paid CAC hiding as an $80 blended number once free organic wins are averaged in, so you keep funding a paid channel that is quietly underwater. Track cost per activated user by channel where you can; the blended average flatters. Second, at a solo operator’s volume you often cannot attribute cleanly at all, which is an argument for channels whose cost is your time rather than your budget, because a misfire there costs a weekend. This is the part I can speak to from inside, at the small scale where the mechanics are visible and nothing is inflating them. Okane Land runs four of these channels itself: SEO written to be cited by AI answers /study/aeo-and-geo-what-the-research-says/ as much as ranked, a self-owned newsletter off an organically built list, a daily reply loop on a sub-200-follower X account that leads with a number and keeps the link out of the main post, and a community forum that the articles feed into. What that daily practice teaches is exactly what the table says: the reply that helps converts better than the post that broadcasts, the list is where everything should point, and the whole thing runs on time, with no ad budget anywhere. Those are mechanics I can show. Any CAC figure I have not measured, I will not claim. Where a one-person AI product should actually spend The verdict is an order of operations. Build the owned email list first, because it converts an order of magnitude better than reach and it is where every other channel should terminate. A follower you cannot email is top-of-funnel; a subscriber is the funnel’s floor. Use community, founder-led building, and cold outreach as feeders into that list, and treat a launch platform like Product Hunt as one-time awareness that mostly rewards people who already have a list to mobilize. Turn on referral once you have a base of retained, happy users, since it multiplies what exists and cannot start from cold. Keep paid off until a channel already converts and the product’s LTV clears the payback bar. For most sub-$50-a-month AI products, that means paid stays off until retention is fixed, because paid acquisition into a leaking product just empties the bucket faster and with a receipt. Which lands on the through-line the channel tables leave out. The cheapest customer is the one your own time can acquire repeatably, and the real work is making the product retain, so that acquiring a customer is worth paying for at all. Fix the leak first. Then the channel you pick barely matters, and only then can you afford to buy your time back with paid. One email, when there's something worth sending Get the research in your inbox. No fixed schedule, no filler. You get an email when we've tested something, run the numbers, or found a tool worth your time. Free. Double opt-in, unsubscribe in one click. What is your real CAC? Compare notes in the forum ↗ https://community.okaneland.com Sources & how we researched this - WordStream / LocaliQ 2025 , Google Ads and Facebook Ads benchmarks. wordstream.com - Unbounce 2025 , Conversion Benchmark Report SaaS landing pages . unbounce.com - HockeyStack 2025 , LinkedIn Ads Benchmark Report. hockeystack.com - First Page Sage 2026 , Customer Acquisition Cost by Channel agency data; sells organic . firstpagesage.com - Ahrefs 2026 , AI Overviews reduce clicks by 58% update . ahrefs.com - Pew Research Center 2025 , Google users click less when an AI summary appears. pewresearch.org - SparkToro / Datos 2024 , Zero-click search study. sparktoro.com - RevenueCat 2026 , State of Subscription Apps AI apps churn ~30% faster . revenuecat.com - Growth Unhinged Kyle Poyar x ChartMogul 2025 , AI retention by price tier. growthunhinged.com - ICONIQ AI gross-margin data, via SaaStr / Tanay Jaipuria 2025-2026 . - Schmitt, Skiera & Van den Bulte 2011 , Referral Programs and Customer Value, Journal of Marketing Wharton . - David Skok, For Entrepreneurs 2013 , SaaS Metrics 2.0 origin of LTV:CAC 3:1 and payback . - Grow & Convert, The CAC of Content Marketing labor-cost sensitivity . growandconvert.com - Hustle Fund, Why you should never average your CAC. hustlefund.vc - Freemius 2025 , State of Micro-SaaS citing MicroConf survey . freemius.com - OpenView / Amplitude 2022 , Product Benchmarks PLG funnel . openviewpartners.com