Paid search is still the fastest way to buy American customers, and the two weeks of spend it takes to prove it on your own account

Paid search is still the fastest way to put an American buyer in front of your offer, and it is the only channel where you can know inside fourteen days whether it works for you. The decision is not whether paid search works. It is whether it works at your price, and the only honest way to settle that is a bounded test with a figure attached to it before you start. Budget 2,000 to 4,000 dollars of media across fourteen days on exact and phrase match terms that name what you sell, send every click to one page you control, and count booked conversations rather than clicks. If the cost per booked conversation lands under a third of your gross margin per customer, scale it. If it does not, you have bought an answer for the price of a small campaign instead of a quarter of guessing.

The fourteen day test, costed

Everything below is arithmetic on your own inputs, which is the only kind of number worth walking into a budget meeting with.

Line	What to commit	What it buys you
Media	2,000 to 4,000 dollars over 14 days, one market, one offer	Enough clicks for a cost per lead you can read, not a cost per lead you have to believe
Landing page	One page, one form, one phone number	Removes the site as a variable, so a bad result is the ad or the term and not the navigation
Call tracking	A tracked number on that page only	In most American service categories the phone outperforms the form, and an untracked phone call looks like a failed test
Negative terms	Written on day one, reviewed on days 3, 7 and 14	The single highest return hour in a new US account
Match types	Exact and phrase only for the test	Broad match on a new account spends the budget learning your category rather than selling your product
Decision rule	Written down before day one	Stops the test being extended indefinitely because the answer was not the one you wanted

Where an American account leaks before the click is a buyer

Three leaks account for most of the waste we see described in US accounts, and all three are settled by a decision rather than by a tool.

• The brand term argument. Bidding on your own name is cheap and converts, which makes it look like the best line in the account. It is a defensive line, not a growth line. Report it separately or it will flatter every other number beside it.
• Geography written for a sales territory instead of a drive time. A state level radius in a metro like Chicago or Dallas buys clicks from people who will never travel to you. Set the radius to the distance a customer actually comes from, then widen it only when the tighter one runs out of volume.
• Automated bidding switched on before there is conversion data to bid against. Smart bidding optimises toward the signal you give it. Give it form fills when your revenue comes from phone calls and it will buy you form fills.

What the FTC rules change about the ad itself

Search ads in the United States sit under Section 5 of the FTC Act, which makes unfair or deceptive acts or practices unlawful, and the agency reads an ad as a whole rather than clause by clause. Three practical consequences for a paid search account.

• A price in an ad has to be the price a buyer can actually get. A headline figure that turns out to need a twelve month commitment belongs with that condition disclosed close to it, not in a footer.
• Claims that depend on a typical result need substantiation you hold before the ad runs, which in practice means your own records rather than a vendor case study.
• If you run testimonials or reviews in ad extensions or on the landing page, the FTC Endorsement Guides apply to them, and the Commission's 2024 rule on consumer reviews and testimonials made the use of fake or materially misrepresented reviews directly actionable with civil penalties. The practical test is whether a reasonable buyer would understand the relationship behind the review.

None of that slows an account down. It changes what goes in the ad copy, and it is far cheaper to settle on day one than to rewrite a whole account after a complaint.

What the assistants said when American buyers asked this question

On 22 October 2026 we put a 198 question American buyer bank to the ChatGPT and Gemini product surfaces through a data provider, and we counted what came back. The market was fixed inside the text of each question, because the provider accepts gl, hl and location parameters and ignores all of them, so the only reliable way to ask an American question is to say so in the question.

Excluding the brand batch, which names us in the question and is therefore a ceiling rather than a measurement, there were 64 American readings across the two engines. Across them the engines produced 2,866 citation URLs on 1,143 distinct hosts. Thirty eight of those 64 readings named no supplier at all, which is the largest single outcome in the American set and is worth more to a buyer than it first looks. On a question about whether paid search still works, an answer that names nobody is an answer that hands the decision straight back to you.

The two engines behaved differently enough that averaging them would destroy the finding. ChatGPT ran a live search on all 32 of its American readings. Gemini ran one on 16 of its 32. So on an American question, one of these two surfaces is reading the live web every time and the other is reading it about half the time, and a page published today reaches them on different schedules.

Where a supplier was named, 53 namings went to software rather than to agencies. Semrush was named in 14 readings, HubSpot in 12, Ahrefs in 9, BrightLocal in 5. Not one American agency was named anywhere in the blind American pass. If you sell marketing services in the United States, that is the gap in one sentence.

What to do in the next two weeks

• Write the decision rule and the budget cap on one line, and get whoever owns the money to agree to it before the first ad goes live.
• Build the negative keyword list before the campaign, from your own sales notes, not after the first invoice.
• Put a tracked phone number and a form on one page and nothing else on it.
• Read the account on days 3, 7 and 14 and change one thing at each reading, so you can tell what caused what.
• On day 14, make the call you wrote down on day zero.

Paid search answers fastest. It does not answer cheapest, and the comparison that matters to most American businesses is paid search against organic visibility on the same monthly figure, which we costed separately in three campaigns, a tight radius and a budget you can defend. If your category is one of the expensive American auctions, the metro arithmetic in the most expensive auction in the country will change the figure you commit.

Questions American advertisers ask about speed and cost

How fast is fast, in days?

A new campaign can serve impressions within hours of approval. A readable cost per booked conversation takes about fourteen days in most American service categories, because that is roughly how long it takes to accumulate enough conversions to separate signal from noise on a two to four thousand dollar test.

Should I test Google and Microsoft Advertising at once?

Test Google first on its own. Microsoft Advertising is worth adding once the terms and the offer are settled, because its lower volume makes it a poor instrument for finding out whether the offer works at all.

Is 2,000 dollars enough for the test?

In most categories, yes, for one market and one offer. If your category's clicks run above 20 dollars, which several American legal, insurance and medical terms do, raise the budget rather than extend the calendar, because a thin test stretched over six weeks mixes in seasonality.

What if the test works and then stops working?

Re-read the three leaks above in order, starting with the negative list and the radius, because a campaign that worked and then stopped has almost always widened its own matching. Check the search terms report before you touch bids.

Does any of this affect whether an assistant names us?

Paid search and assistant answers are different instruments and they do not substitute for each other. Paid search buys a click today. Being named in an assistant answer depends on what the engines can read about you, and in 64 American readings on 22 October 2026 the engines leaned on software vendors and documentation sites rather than on advertisers.

The answer a buyer gets when they ask an assistant instead of Google

An American buyer who asks ChatGPT or Gemini whether to run ads is not shown your ad, and in 38 of our 64 American readings they were not shown a supplier either. That is a second queue, with a different gatekeeper, and the cost of entry is whether the engines can read a clear, dated, checkable answer about you somewhere they already look. AI Knows Us reads ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews and records which of them names you and which sources they used to decide, which is the measurement that sits beside your ad account rather than inside it. In the United States that sits on the 299 dollar a month Growth plan, the cheapest plan that reads all five engines, read on aiknowsus.com/pricing on 22 October 2026.