Story

Are Paid Ads Worth It for Indie Hackers? The 'Slot Machine' Debate

Are paid ads worth it for a product that has to earn the click back on a single order? On 11 November 2026, Global Commerce Graph published “The 98% Slot Machine: How Paid Advertising Became an Extraction Tax”. The opening example is a founder who spends $420, gets one sales call and one small order, and makes roughly $410. The line under it is “A net loss of ten dollars.” The essay presents that scene as an illustration.

The core claim is that as more businesses bid, prices rise “to the absolute maximum point of economic tolerance: the exact dollar amount where the advertiser is breaking even on their gross margin.” The essay’s line for that advertiser is: “They are playing a digital slot machine calibrated by hyperscalers to pay out at exactly 98%.” The 98% is a metaphor. The essay presents no data for it. It is a vendor essay. The piece ends by pitching Global Commerce Graph’s own product, a graph of companies under 100 employees that surfaces target buyers.

This was written on 20 November 2026. The Hacker News thread was submitted on 19 November 2026 at 18:35 UTC. When checked, it stood at about 91 points and about 45 comments, and it was on the front page on the morning of 20 November, AEDT. Counts on that thread move. Founder figures below are self-reported. One click-cost series is 12 accounts.

Are paid ads worth it, or a slot machine?

stevage wrote: “They do not present any data at all to support their 98% figure.” uptrails claimed the article is AI-generated. VBprogrammer disagreed: “It has a voice and a thread running through it which I’ve never seen an LLM manage.” Those are two different objections. The missing table is the one that touches the 98%.

Other comments describe an auction, and they leave room for both sides to gain. cm2012 wrote: “It’s an auction. It’s not a conspiracy. Every marketing channel has its pros and cons.” PotenRoyal wrote: “Advertising is not a zero sum game. There is room for both sides to walk away with gains from trade.”

nkmnz wrote: “in economic theory, perfect competition drives the margin to zero.” PaulHoule answered with the boundary on that claim: “The margin for the marginal participant! The more profitable participants will still make a profit.” danielmarkbruce wrote: “most players using paid advertising can’t win.” davidfischer wrote: “The best advertisers are setting the floor prices for ads within their niche.”

mangoceylon, who says they work in ads, wrote: “A way bigger factor than anything mentioned in this article is simply what you as an advertiser can afford as a CAC”. CAC is what it costs you to acquire a customer. dzink wrote: “The only way to keep out of those traps is to have real metrics of your own and benchmarks of your own.”

A paid click is one way to buy a visit. Sponsor slots, bid-to-rank boards, and performance ads are three different purchases, and a click from Google sits in the third. Whether you need an audience before a buyer can find the product is the neighbouring question, about distribution when you do not already have a crowd.

Why are Google Ads getting more expensive?

On 19 November 2026, Danny Goodwin at Search Engine Land summarised a New York Times report. Some longtime advertisers are cutting Google search ad spend as costs rise and AI Overviews push ads down. What an AI Overview does to the unpaid click is a separate measurement, from a randomized study of outbound organic clicks. The Search Engine Land piece is about the paid listings on that same page, and about advertisers spending less on them.

Marshall Evans, a North Carolina marketer for small businesses, described one trophy seller in that report. Monthly Google search spend fell 75%, from $2,000 to $500. The seller shifted spend to Tubi, Bing, Reddit, and Nextdoor. Evans said: “A.I. takes over such a large amount of the search results now”.

Max Anderson, CEO of OnlyFinder, said: “Our costs have probably tripled or quadrupled in the last three years”. He cut Google Ads spend to roughly half. That change in cost, and the cut, are his account.

Google’s statement in the same report is that AI features drive more searches, including commercial queries. Sundar Pichai said: “We are focused on building the best search user experience, followed by monetization”.

Conner Crowe updated “Are Google Ads Clicks Getting More Expensive? (2023 to 2026 Data)” on 30 October 2026. The set is 12 local businesses whose ads he manages, mostly healthcare and home services. It is a small sample, and the accounts are his.

Eleven of the 12 pay more per click than in summer 2023. A typical click costs about 2.27 times the summer 2023 price. $1,000 that bought 100 clicks in summer 2023 bought about 44 in summer 2026. The typical cost per click rose 19% in 2024, 11% in 2025, and 76% in 2026. Ten accounts that had not advertised every summer paid about the same per click this year as last year.

When should a SaaS start running paid ads?

The comments that come closest to an answer use a number the advertiser already has. mangoceylon’s factor is the CAC you can afford. dzink’s condition is metrics of your own. PaulHoule’s reply to the zero-margin point is that it is the margin for the marginal participant, and that more profitable participants will still make a profit. davidfischer’s floor is the price the best advertisers in the niche are setting.

Replit’s coding agent can run Google Ads, announced on 12 November 2026. Sam Whitaker’s Indie Hackers report the next day says the spend comes out of Replit credits. The minimum average daily budget is $15. The help page, as that report gives it, says the budget “is not a hard daily spending cap”. Campaigns are Performance Max only, and one at a time. The same page says: “Do not assume the campaign will stop automatically.”

Replit says entrepreneurs connecting Replit-built businesses to Stripe grew ninefold since May. That multiple is company-reported, and it counts Stripe connections. The agent that can buy the clicks is a coding agent. What indie hackers now budget for AI coding tokens is the other bill on that kind of tool. A product you own, a price, and a buyer you can reach is the starting condition already on the record. The $15 is the minimum average daily budget in Replit’s setup.

How do you know if your ads pay back?

dzink’s sentence is the thread’s test: real metrics of your own, and benchmarks of your own. motionlessveloc quoted Wanamaker: “Half the money I spend on advertising is wasted; the trouble is I don’t know which half.” The essay’s $420 and roughly $410 are the illustration from the opening. A net loss of ten dollars, in that scene, is the essay’s arithmetic. It is an example, and the essay does not present it as data.

The window you measure in can move the result. On 12 November 2026, lukehoang posted on Indie Hackers. He discloses that he works at Airbridge, which does mobile attribution. The App Store bills a free trial in its last 24 hours. Meta’s default counts conversions up to 7 days after a click. With a 7-day trial, the first payment lands at the edge of that window, so return on ad spend can look worse than the cohort.

His example is made up. Return on ad spend reads 0.4 while the cohort is closer to 1.2. The advice in the post is to optimise for the trial start, judge the spend on cost per trial, and check the cohort at day 14 and at day 30. Whether a small product still gets paid depends on a customer after the click, which is a later figure than a 7-day return.

How do you stop click fraud on Google Ads?

On 14 November 2026, Sam Whitaker reported a check built by Ilias Ism, who runs 24 sites. Ism says the check caught 14 bot clicks worth $206 in one morning. That figure is self-reported. He matched server logs against the Google Ads API, blocked 30 subnets, and filed a refund claim. On the off-the-shelf tools, he wrote that “$149/mo fraud click tools can’t even see” headless bots.

The steps named in that report are a request through Google’s Click Quality refund form, and blocks. Google allows 500 IP blocks per campaign. A commenter advised checking Google’s own Invalid clicks column first, because filtered clicks are not billed. The $206 is one morning, on his sites, as he reported it.

Two months to $500 MRR, and no ad spend

On 15 November 2026, rodvaN wrote that ViralHunt.io had reached $500 in monthly recurring revenue in two months. That figure is self-reported. The post says: “I haven’t spent a single dollar on PPC or ads. Everything has come through ChatGPT and Claude recommendations.”

The month is one product, posted by the person who runs it. He names ChatGPT and Claude recommendations as the source of that revenue. What the first $1,000 in monthly recurring revenue does and does not mean is the public milestone this $500 is halfway toward.

A short timeline

When What is on the record
30 October 2026 Crowe updates click costs for 12 local accounts he manages. Small sample. Eleven of 12 cost more per click than in summer 2023.
11 November 2026 Global Commerce Graph publishes the slot-machine essay. The 98% is a metaphor. The essay presents no data for it, and it ends on the company's own product.
12 November 2026 Replit's agent can run Google Ads. lukehoang posts on a 7-day trial and Meta's 7-day click window. His 0.4 and 1.2 are a made-up example.
13 November 2026 Whitaker reports the Replit setup: credits, a $15 minimum average daily budget, Performance Max, one campaign, and a budget that is not a hard cap.
14 November 2026 Ism's homemade check, self-reported: 14 bot clicks, $206, one morning. He blocked 30 subnets and filed a refund claim.
15 November 2026 ViralHunt.io at $500 MRR in two months, self-reported, with no PPC or ads.
19 November 2026 The Hacker News thread goes up at 18:35 UTC. Goodwin's Search Engine Land piece summarises advertisers cutting Google search spend.
20 November 2026 The thread is on the Hacker News front page in the morning, AEDT. About 91 points and about 45 comments when checked.

The essay is The 98% Slot Machine. The thread is on Hacker News. Crowe's series is on his site. The advertiser cuts are in Search Engine Land. Replit's agent is in Whitaker's 13 November report. The trial window is lukehoang's post. The fraud check is Whitaker's 14 November report. ViralHunt is rodvaN's post.

The 98% has no table under it. The auction comments describe a price set by other bidders, with a profit left for the advertisers who are not the marginal ones. Crowe’s 12 accounts, mostly healthcare and home services, paid more per click than in summer 2023, and ten accounts that had not advertised every summer paid about the same as last year. A coding agent can open a Google Ads campaign on Replit credits, and the help page says not to assume the campaign will stop. One founder matched bot clicks to his own logs. Another posted $500 in monthly revenue and said none of it came from ads.

What do you think? Have paid ads ever paid back for you?

FAQ

Why are Google Ads getting more expensive?

On 19 November 2026, Danny Goodwin at Search Engine Land summarised a New York Times report that some longtime advertisers are cutting Google search ad spend as costs rise and AI Overviews push ads down. Marshall Evans, a North Carolina marketer for small businesses, described one trophy seller whose monthly Google search spend fell 75%, from $2,000 to $500. That seller shifted to Tubi, Bing, Reddit, and Nextdoor. He said A.I. takes over such a large amount of the search results now. Max Anderson, CEO of OnlyFinder, said their costs have probably tripled or quadrupled in the last three years, and he cut Google Ads spend to roughly half. That cost change is his account. Google says its AI features drive more searches, including commercial queries. Sundar Pichai said the company is focused on building the best search user experience, followed by monetization. A separate series, from Conner Crowe on 30 October 2026, covers 12 local businesses whose ads he manages, mostly healthcare and home services. It is a small sample of his own accounts. Eleven of the 12 pay more per click than in summer 2023. A typical click costs about 2.27 times the 2023 price. $1,000 that bought 100 clicks in summer 2023 bought about 44 in summer 2026. The typical cost per click rose 19% in 2024, 11% in 2025, and 76% in 2026. Ten accounts that had not advertised every summer paid about the same per click this year as last year.

When should a SaaS start running paid ads?

The Hacker News comments on the slot-machine essay answer with a number from your own revenue. mangoceylon, who says they work in ads, wrote that a bigger factor than the essay is what you as an advertiser can afford as a CAC. dzink wrote that the way out of the trap is real metrics of your own and benchmarks of your own. PaulHoule wrote that the margin in question is the margin for the marginal participant, and that more profitable participants will still make a profit. davidfischer wrote that the best advertisers are setting the floor prices for ads within their niche. Replit's coding agent, announced on 12 November 2026, can run Google Ads paid from Replit credits. Sam Whitaker reported that the next day on Indie Hackers. The minimum average daily budget is $15. The help page says that budget is not a hard daily spending cap. Campaigns are Performance Max only, one at a time, and the page says not to assume the campaign will stop automatically. Replit says entrepreneurs connecting Replit-built businesses to Stripe grew ninefold since May. That multiple is company-reported. The $15 is a minimum in that product. It does not say a new SaaS has the margin to bid.

How do you know if your ads pay back?

On the Hacker News thread, dzink wrote that the way to stay out of the trap is real metrics of your own and benchmarks of your own. The essay's opening spends $420, books one sales call and one small order, and makes roughly $410, which it calls a net loss of ten dollars. That scene is an illustration, not a payback study. motionlessveloc quoted Wanamaker: half the money spent on advertising is wasted, and the trouble is not knowing which half. On 12 November 2026, lukehoang, who discloses that he works at Airbridge on mobile attribution, wrote that the App Store bills a free trial in its last 24 hours and that Meta's default counts conversions up to 7 days after a click. With a 7-day trial, the first payment lands at the edge of the window, so return on ad spend can look worse. His example is made up. The return reads 0.4 while the cohort is closer to 1.2. He says to optimise for the trial start, judge the spend on cost per trial, and check the cohort at day 14 and at day 30.

How do you stop click fraud on Google Ads?

Sam Whitaker's Indie Hackers report on 14 November 2026 describes a homemade check by Ilias Ism, who runs 24 sites. Ism says the check caught 14 bot clicks worth $206 in one morning. That figure is self-reported. He matched server logs against the Google Ads API, blocked 30 subnets, and filed a refund claim. He wrote that $149/mo fraud click tools can't even see headless bots. The steps named in the report include Google's Click Quality refund form. Google allows 500 IP blocks per campaign. A commenter advised checking Google's own Invalid clicks column first, because filtered clicks are not billed.