Will AI Take Your Next Business Idea? How to Tell

Most business ideas carry an AI question, and most of what you can read about it is either panic or sales copy. There is a third option: score the idea on five questions in one afternoon and get a zero-to-ten exposure number with bands that tell you what to do with it. That is the method this site runs on every course before publishing it, and the Will AI Eat This Business guide teaches it end to end. The short version is below, plus the verdicts where the answer came back ugly.

One framing fact to set the stakes. In the first four months of 2026, 68 percent of US Google searches ended without a click, up from 60 percent in 2024 and 49 percent in 2019, per SparkToro. When an AI Overview appears, position-one click-through drops by roughly 58 percent, per an Ahrefs study of 300,000 keywords. Attention is being intercepted at the search layer. That matters enormously for some businesses and barely at all for others, and the five questions are how you tell which.

The five questions

Score the beginner's version of the business, not the master's. The machine eats floors before it eats ceilings, and you are deciding whether to enter.

First, the deliverable. Is the customer paying for bytes, a physical outcome, or accountability? Pure bytes adds 2 to your score. A bundle that is mostly outcome or accountability subtracts 2.

Second, the entry floor. Could a beginner armed with free AI tools produce a version of the average provider's offer that average customers would accept? If yes, add 2. This is the question that separates businesses that are merely changed from businesses whose first rung has been removed.

Third, the demand gate. Who controls whether customers find you? An ungoverned platform that does not police machine-generated supply adds 2. A policed platform adds 1. Referrals, your own list, or channels where humans are the point subtract 1.

Fourth, the blame. When the work goes wrong, must a licensed, insured, physically present human absorb it? Subtract 2 if yes, 1 if the accountability is real but partial, nothing if customers just tolerate a worse output and move on.

Fifth, the task split. What share of first-year hours sits on tasks a machine could do without you? Over half adds 2. A quarter to half adds 1. Under a quarter holds. Note the asymmetry: overhead automation never lowers exposure much, because every business gets that subsidy. Only deliverable automation raises it much.

Start at 4, the agnostic middle, apply the adjustments, clamp to 0 through 10.

The bands, with real scores attached

Zero to 3 means build it. The machine barely touches the value. Eleven businesses in this catalog scored 2: cleaning, lawn care, pressure washing, handyman, junk removal, laundry pickup, micro bakery, pet waste removal, party rental, notary loan signing, and Airbnb cleaning. The pattern is uniform: a physical outcome produced at a customer's property, wrapped in local trust, often under a rule structure a human must satisfy. No model picks up a duffel. No algorithm notarizes a loan package.

Four to 6 means build it with adaptation. Sixteen courses live there, from appliance repair at 3 through video editing at 3, podcast production at 4, YouTube and KDP at 5, up to print on demand and the social media agency course at 6. The machine does real work inside the value chain, sometimes half of it, and a durable human layer remains: accountability, taste, a relationship, a license, owned demand. The adaptation is a teachable craft, which is why those courses exist.

Seven to 8 means closed for beginners. The entry lane, the simple work that used to train every newcomer and pay the bills while they climbed, is now the machine's lane. Experts may still thrive for years, but a beginner course here would sell a ticket to compete directly with the tool that closed the lane. Nine to 10 means the deliverable is the machine, full stop.

This site draws its line at 7 and enforces it. The 2026 catalog review researched fourteen business categories, planned 41 courses, and built 34. Three were removed for AI exposure, by name: copywriting, which scored 8, the highest score assigned to any planned course; niche websites; and Etsy shop. Four more were removed for capital requirements. The removals are disclosed because a method is only as good as its willingness to lose arguments.

What the evidence says about who gets eaten

The sharp damage so far sits exactly where the rubric predicts: bytes delivered through ungoverned platforms.

On Etsy, a UK illustrator selling for a decade watched sales fall 98 percent in a year as AI portraits flooded listings at a fraction of the price, per WIRED. Another seller who made more than $40,000 in 2021 saw sales drop 30 percent, then 50 percent, and left in January 2025. The labor market matches. A Stanford Digital Economy Lab study found entry-level hiring in AI-exposed occupations declined 13 percent relative to less-exposed ones, concentrated among workers aged 22 to 25 in software development, customer service, and clerical work, a finding the authors themselves call observational. On freelance platforms, the Hui, Reshef, and Zhou study in Organization Science found writing-related jobs down about 2 percent with earnings down 5.2 percent after ChatGPT, and image-related work down 3.7 percent in jobs and 9.4 percent in income.

The platforms are responding, which is the demand-gate question doing its job. YouTube renamed its repetitious-content policy to "inauthentic content" in July 2025 and bars mass-produced video from monetization. Amazon capped KDP uploads at three new titles a day and requires AI-content disclosure. TikTok requires labeling realistic AI-generated content. Policed platforms add 1 instead of 2 for a reason: enforcement is the difference between a lane that exists and one that flooded.

Now the other side of the rubric, because the fear is often misdirected. Run the five questions on a home organizing service and the deliverable is a physical outcome, blame lands on a present human, demand flows through referrals: the guide's worked example scores it 0 or 1. Run them on a resume-writing service and the deliverable is pure bytes, a beginner with free tools matches the average offer, demand runs through platforms: it scores 9. Same economy, opposite verdicts, and both scores came from the same ten minutes of questions.

Scores are snapshots, not prophecies. Re-run them when a platform changes its rules, when a model starts doing your deliverable acceptably, or when prices in the niche start sagging.

What to do with a 5 or a 6

A 5 is not a no. It is a build-with-deflection. KDP at 5 means the deflection is niche research and craft, because AI can write books, which is precisely why Amazon caps uploads and requires disclosure. Print on demand at 6 means anyone can generate a thousand designs a day and everyone does, so the deflection is niche selection and trademark discipline. The video-first and audio-first creator businesses sit at 5 and 4 respectively, held up by platform policing and audience relationships rather than by the deliverable itself.

The adaptation moves are consistent: own the demand channel instead of renting it, sell the judgment layer instead of the production layer, and pick niches where the human is the point. The guide builds that deflection on purpose rather than hoping.

If you want the method with the worksheets, the guide is the full version. If you want a shortcut to ideas that already cleared the bar, the fit quiz only recommends rated courses, and nothing rated 7 or above is taught here at all.