The intersection
Most AI marketing is a content generator with a subscription. The thing that actually changed is who answers the question first.
Your buyer no longer starts at a list of ten blue links. They ask an assistant, get one answer with three or four sources, and form a preference before any website is visited. That is a different problem from ranking, and almost nothing sold as AI marketing addresses it.
Written by Kristian Baer, Northtek · Anchorage, Alaska · Updated 2026-08-21
The short answer
What does an AI marketing agency actually do differently?
Two things, only one of which most firms sell. The visible one is production: AI writes drafts, edits video, generates variants. That is real efficiency and it is also available to every competitor, so it is not an advantage. The one that matters is discovery. When a buyer asks an assistant for a recommendation, the assistant resolves entities and assembles an answer from sources it can attribute - so the work is making your organisation a resolvable entity with corroborating evidence, and publishing information that exists nowhere else so there is a reason to cite you rather than a competitor. Content that restates what every other firm says has no reason to be selected, no matter how much of it there is.
What changed
Ranking and being cited are now two different problems.
Ranking is about being one of ten results a person then chooses between. Citation is about being one of three sources an assistant uses to compose the only answer the person sees. The second is a much smaller doorway.
What gets through it is not volume. It is information gain - content containing something that does not exist elsewhere. An assistant assembling an answer has no reason to pull from the fifth article restating the same advice. It has every reason to pull from the only organisation that measured something.
That reframes the entire content question. The winning move stops being "publish more about the topic" and becomes "publish the thing nobody else can", which is usually a measurement, an original dataset, or specific operational knowledge earned by doing the work.
Where AI genuinely helps
Three real gains and one honest warning.
The production tools are useful. They are just not a differentiator, because everyone has them.
- Speed on the mechanical half
- Drafts, variants, transcripts, cut-downs, alt text, translation. Genuine time savings on work that was always mechanical, and the right place to use it.
- Response time as a conversion lever
- Answering an inbound inquiry in minutes rather than hours moves conversion more than most creative decisions. That is an operations problem wearing a marketing hat, and it automates well.
- Entity infrastructure
- Structured data declaring a named author and organisation, corroborated by external profiles. Unglamorous, mostly invisible, and the thing that determines whether published work accrues to anything.
- The warning
- AI-written content published at volume is the fastest way to become indistinguishable. If a competitor could have published it, it will not differentiate you, and the machine that writes it also cannot tell you what is true about your business that nobody else knows.
Why us for this specifically
We run the experiment on ourselves and publish the numbers.
This site is the demonstration. It declares a named founder in structured data on every page, links every external profile that corroborates it, and publishes original research rather than commentary. It also carries fourteen industry playbooks written for the person doing the work rather than for a keyword.
We measure it too. When we found that twenty percent of the body text on our own new pages repeated across them, we rewrote all of it and got that to under two percent, because templated content is exactly what an assistant has no reason to cite. Most agencies do not measure that on their own site, let alone yours.
The unusual part is that the same building does the engineering. An agency that only markets can tell you entity declarations matter. Whether they can implement the structured data, fix the duplication, and build the thing worth citing is a different question.
Apply it
What we would do first for an Alaska business
In this order, because each step makes the next one cheaper.
-
01
Fix the entity declarations
A named human in the structured data, an author byline on substantive pages, and every external profile linked. Roughly a day of work, and content published before it is done does not compound.
-
02
Find the thing only you can say
The operational knowledge earned by doing the work. A number nobody else measured. If a competitor could have written it, it will not get cited.
-
03
Publish where your buyer already reads
For most Alaska businesses that is a trade association newsletter and LinkedIn, not a company blog. One column in front of the right five hundred people beats ten posts in front of nobody.
-
04
Measure what an assistant says about you
Ask the assistants your buyers use, monthly, and write down the answer. It is crude and it is the only direct read on whether any of this is working.
Straight answers
Is SEO dead?+
No, and anyone saying so is selling something. Assistants draw heavily on the same index, so technical health, crawlability and genuine authority still decide whether you are eligible to be cited at all. What changed is that being eligible is no longer sufficient - among eligible sources, the one saying something distinctive gets selected.
Should we just publish more content?+
Only if it is content nobody else could publish. Volume of restated advice is now actively counterproductive: it costs money, dilutes the site, and gives assistants nothing to prefer you for. Twenty pages that say something beat two hundred that do not.
Do you use AI to write client content?+
For drafts, structure and variants, yes, and we would be lying if we said otherwise. What it cannot do is know what is true about your business that nobody else knows, and that is the part that determines whether the content works. The interview to extract that is the expensive part, and it is a human one.
How long before this shows results?+
Entity and technical work takes effect as pages are recrawled - weeks. Authority and citation are quarters, not weeks. Anyone promising faster is describing paid reach, which is a legitimate service and a different one.
We are a small business. Is this relevant?+
The entity work is more relevant at small scale, not less, because a small business is easy for an assistant to fail to resolve at all. It is also cheap - a day of structured data and profile linking, once. The research and publishing side scales with ambition.
Can you do the marketing without the AI engineering?+
Yes. Plenty of clients use one side only, and we scope what you asked for. See SEO, web design and content for the marketing work on its own.