Best way to advertise a small business compared, ranked by what each channel costs to test

For a small business the useful ranking is not which channel is best. It is which channel gives you a readable answer for the least money, because you cannot afford to be wrong twice.

Ranked by cost to get a real answer, cheapest first

One, your existing customers. Costs nothing. Email or phone the last hundred people who bought from you with one specific offer. If that does not work, you have learned something enormous about your offer for free, and no paid channel was going to rescue it. Almost nobody does this first and it outperforms everything below it.

Two, your Google Business Profile. Costs nothing. If you serve an area, this is the highest intent free placement available to you. Complete it, get the categories right, collect reviews. People searching for what you do near where you are will see it. A profile left half finished is the most common free money left on the table in American small business marketing.

Three, the one directory or marketplace your customers actually use. Varies by trade and it is usually obvious once you ask two customers how they found you. Cheap, and it either works quickly or it does not.

Four, paid search. This is where it gets interesting, because the test cost is set by your category's click price and nothing else. Google publishes no minimum. We read business.google.com/us/google-ads/ on 21 October 2026, reached by a 301 redirect from the old ads.google.com home page, and the only commitment language on it is that "New advertisers will receive an ad credit after meeting the minimum spend requirement for the selected offer". The requirement is not stated, and the page's own offer amounts render as unfilled template placeholders rather than dollar figures. So nobody can tell you the floor from that page, including us, and you have to compute it yourself. The campaign types it names are Performance Max, Search, Shopping, Video Reach, Video View, Demand Gen and App.

Here is how to compute it in two minutes, which is the practical content of this page. Find your likely cost per click. Divide your test budget by it to get clicks. If that is under about two hundred clicks, you will not get a readable conversion answer, you will get noise, and you should either narrow the targeting until the click price falls or not run the test. Two hundred clicks is the number to plan around, not a dollar amount, because the dollar amount is different in every category.

Five, paid social. Cheaper clicks than search and colder intent, so it needs more volume to read. Better at creating demand than capturing it, which means judging it on cost per click will mislead you.

Six, content networks. Reach is not your constraint. Taboola's advertiser page, read at taboola.com/advertise on 21 October 2026, describes its product as "Performance AI with a global network of vetted publishers, reaching 600M daily active users" and publishes no minimum spend. For a small business the constraint is having something worth sending six hundred million people to, and usually you do not yet.

Seven, print, radio, outdoor and sponsorship. Not useless, genuinely unmeasurable at small scale. Buy these when you have a working measurable channel and want brand coverage on top, never as the first test.

The three mistakes that waste the most small budgets

Running three channels at once. With a small budget, three channels means three unreadable tests. Run one. Read it. Then run the next.

Stopping at four weeks. Most small advertising is stopped before it produced enough data to judge, and then the next channel gets the same treatment. Decide the number of clicks or conversations you need before you start, and let it run to that.

Spending on reach before fixing the landing. If people arrive and cannot see a price, a phone number or a reason to trust you, more arrivals is a more expensive version of the same result.

The channel you cannot advertise on, and it is now a real one

Here is something a small business owner should know that is on none of the lists above, because there is nothing to buy on it.

A customer asks ChatGPT, Gemini, Perplexity, Claude or Google AI Overviews who to use for what you do, in your area. The assistant names two or three businesses. There is no advertising slot in that answer. You cannot bid for it, you cannot sponsor it and there is no account to open. If you are not named, the customer never arrives and nothing in any of your channels records a customer who never arrived.

For a small local business this matters more than it does for a large one, because two or three names is the whole answer, and in a local category there may only be a handful of credible businesses to choose from.

AI Knows Us measures this one thing. We are not an advertising service. We do not place ads, we do not manage budgets and we are not an alternative to anything in the seven above. We take the buying questions your customers ask and put them to all five surfaces, then report which businesses get named.

And we will be straight about the price, because this page is about small budgets. Our ladder at aiknowsus.com/llms.txt, read 21 October 2026, publishes Growth at $299 a month as the cheapest plan reading all five surfaces, with 100 buyer questions watched and 1,600 credits. For a very small business, $299 a month is a real decision and it competes directly with the paid search test above. If you have not yet done steps one and two on this page, which are free, do those first. SEO Start at $149 publishes 0 buyer questions, 0 credits and 0 engines and does not read answer engines, so the $149 figure is not a cheaper way in and we will not let it look like one. You can also simply ask the five assistants yourself once, by hand, for nothing. That is a single snapshot rather than a monitored programme, and for many small businesses a snapshot is the right starting amount of information.

What to read next

For the general version of this ranking, covering businesses of any size and more channels, see best ways to advertise your business, ranked by what each channel costs to test. We link to it deliberately. When we measured inbound links across 2,286 stored page bodies on 21 October 2026, no other page on this site links to that one, and 939 of our 2,200 live pages were in the same position, 42.7 percent, an upper bound.