The most underrated discovery channel in the United States right now, and the budget it takes to own it

The most underrated discovery channel for an American business in 2026 is not a platform. It is the set of third party sources that AI assistants cite when they answer a buying question, and the reason it is underrated is that almost nobody has looked at the list. We have. On 22 October 2026, across 64 blind readings of ChatGPT and Gemini on American buying questions, the engines produced 2,866 citations on 1,143 hosts, and the concentration at the top was not vendor marketing. It was documentation, trade press, two agency directories and the regulator. Getting accurately represented on those costs a fraction of what an American business typically spends on paid social, and it feeds the queue nobody is reporting on.

The actual list, with the reading counts

Host	Readings citing it, of 64	What it is
developers.google.com	17	Google's own technical documentation
support.google.com	15	Google's own help properties
searchengineland.com	13	Trade press
blog.google	11	Google's own announcements
clutch.co	11	Agency directory
semrush.com	10	Vendor property
ftc.gov	9	The regulator
blog.hubspot.com	9	Vendor property
brightlocal.com, searchenginejournal.com, business.google.com, developers.google.cn	7 each	Vendor, trade press, Google properties
consumer.ftc.gov, goodfirms.co	6 each	Regulator, agency directory

One entry in the raw host data, ccn.web.tr at 8 readings, is the data provider's own fetch infrastructure rather than a publisher. It is excluded here and should never be presented as a source or merged into another host's count.

What it costs to own the channel

• Directory profiles, done properly. The two directories in that list are free or low cost to list on. The real cost is an afternoon per profile making the service description, the price position and the sector claims match your own site exactly, then a quarterly recheck. Budget four hours a quarter.
• Trade press, earned rather than bought. One piece of original, dated, checkable data about your market is the cheapest route into American trade press. Budget the cost of collecting it, which is usually internal time rather than money.
• Documentation aligned pages. The engines cited Google's own documentation more than anything else. Pages on your site that accurately reflect what those documents actually say, with the date you read them, get quoted because they agree with the source the engine already trusts. Budget writing time.
• Regulator aligned accuracy. ftc.gov and consumer.ftc.gov appeared in 15 readings between them. Pages that state the FTC position correctly, including the Endorsement Guides and the 2024 rule on consumer reviews, are both safer and more quotable. Budget a review step.

Put together, that is a few days of work a quarter and no media spend, against the channel it competes with for budget, which in most American plans is paid social at four figures a month and rising.

Why it stays underrated

Three reasons, all structural. The channel has no sales team, so nobody calls an American marketing director to pitch it. It has no dashboard by default, so it does not appear in the monthly pack. And its results are invisible without a reading, because an assistant answer is not a page view. Those three together mean the work does not get proposed, does not get reported and does not get noticed when it works, which is exactly the profile of an underpriced channel.

The opening in the numbers

In the 64 blind American readings on 22 October 2026, the engines named no supplier at all in 38 of them. That is the largest single outcome in the American set. An engine answering a buying question in generalities has committed to nobody, so the slot is open and taking it means publishing a specific, dated answer rather than displacing an incumbent. Where a supplier was named, there were 53 namings and all of them went to software: Semrush 14, HubSpot 12, Ahrefs 9, BrightLocal 5, and six smaller products once or twice each. No American agency was named once. The engines also behaved differently, with ChatGPT running a live search on all 32 of its American readings and Gemini on 16 of its 32, which is why both are reported separately and never averaged.

How to start this quarter

• Take your own baseline. Twelve buying questions, both engines, United States named in the question text, answers and sources saved with the date.
• Count the hosts in your own source lists. The repeat hosts are your target list, and they will differ from your link building list.
• Fix the directory profiles the engines read first, because it is the cheapest item and the fastest to complete.
• Commission one original dated figure about your own market. That is the asset trade press will take and the engines will quote.
• Re-read in thirty days with the same questions, reporting the two engines separately.

If your business is local, the city level source lists diverge sharply from the national one and the Houston reading in particular shows a newsroom winning, which is set out in a newsroom made the top four. If you want to understand what separates a citation list from a media list before you brief anybody, that distinction is drawn in the citation list is a media list.

Questions about the sources assistants read in the United States

Is this just link building with a new name?

The target list is different, which is the whole point. A link building brief chases domain authority. This chases the specific hosts an engine cited when it answered your buyers' questions, and on our reading those were documentation, trade press, directories and the regulator rather than high authority blogs.

How quickly does a directory fix show up?

Treat the first re-read as thirty days out, and remember that one of the two engines we measured ran a live search on every American reading while the other ran one on half, so the two will reflect a change on different schedules.

Do vendor blogs matter, given semrush.com and blog.hubspot.com appear?

They appear because those vendors publish a great deal of specific, dated material. The lesson is about the shape of what they publish rather than about their domains, and that shape is reproducible by anybody.

What is the minimum viable version of this?

One baseline reading, two directory profiles corrected, and one original dated figure published on your own site. That is a quarter's work for one person at part time and it is a real programme.

How do I know it worked?

By re-reading the same questions on the same engines with the same denominator and comparing the three way score. Named you, named a rival, named nobody. Without a baseline the question has no answer, which is why the baseline comes first.

The channel with no dashboard, and the dashboard for it

This channel goes unfunded because nothing reports it. AI Knows Us reads ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews for the buying questions in a category, records which of them named you and which hosts each answer used, splits the result by engine with its denominator, and keeps the raw answer text so the figure can be audited rather than trusted. In the United States that is the Growth plan at 299 dollars a month with 100 buyer questions watched and 1,600 credits, the cheapest tier reading all five engines, read on aiknowsus.com/pricing and aiknowsus.com/llms.txt on 22 October 2026.