The boring press release machines learned to love

What the results mean for carriers, brokers and 3PLs is not an abstraction. An example: When a shipper asks a large language model (LLM) which provider handles omnichannel distribution out of Florida, the answer arrives before anyone visits a website. Whoever published a number wins and gets named. To those unaware, they missed out on a conversation they never knew was happening.

“AI Cannot Invent a Number”

The finding is narrower than “send more press releases.” Luckily, that narrowness is the point. Ordinary releases built around company announcements, personnel changes or awards generated minimal citations. The releases built around a hard figure carried the whole result.

“AI cannot invent a number, so it cites whoever published one,” said Kara Brown, CEO and co-founder of LeadCoverage, in an interview with FreightWaves.

That constraint explains the mechanics behind the magic. Language models can only generate; they do not report. When a query demands a figure, the model reaches for a source that supplied one, and wire copy is unusually easy to reach. Every release on GlobeNewswire shares the same skeleton: headline, subhead, data, and quotes. AP style, uniformly applied, turns out to be machine-readable by accident.

Brown’s warning to companies sitting on proprietary data doesn’t mince words.

“The companies that publish specific, useful data on a consistent schedule are the ones AI cites most, and that citation is often the first impression a prospect gets before they ever visit your website,” she said. “The companies sitting on their data simply aren’t getting citations, and they never see the potential prospects and deals that pass them by.”

Why AI Citations Favor the Middle Market

Google was always an auction. That is the part freight marketers understood, and the part that priced most of them out.

“If you pay Google money, they will put you at the top of the answer whether or not it’s organic or paid,” Brown said. “If you don’t pay Google, they will diminish your visibility on Google.”

She has a word for the arrangement: mercenary. Google has advertisers to serve and a business reason to serve them. The LLMs, at least for now, run on different incentives. There is no keyword auction on a citation.

For a mid-market 3PL, broker, forwarder or tech vendor with a specific niche, that gap is the entire opportunity, because the traditional route is closed.

“You can’t compete with Old Dominion on LTL,” Brown said. “They already own the search volume for LTL. Trying to outrank them on that term isn’t a fight worth picking.”

The opening has a clock on it. Brown describes the citation effect as a flywheel with a half-life: early participants accumulate weight the way compound interest does, and latecomers spend their budget fighting incumbents who started first.

“The earlier you start, the more time you have to let this half-life percolate with the LLMs,” she said. “The later you start, the more you’re competing with the folks that have already started.”

Money is no longer solving it in a post-search world. The old escape hatch, outspending the field on keywords, does not exist inside an answer engine.

Trade Press, Reweighted

Two figures from Muck Rack are reframing where the effort should go.

About 1% of all answer engine optimization citations come directly from a press release, which works out to roughly 33,000 searches a day resolved by wire copy. Separately, 27% of industry-specific searches are answered by trade publications.

That second number matters more in freight than almost anywhere else, because freight queries are never generic. Nobody asks an LLM for a dentist nearby. They ask which 3PL runs omnichannel distribution near a Florida headquarters, and the model looks for a publication that has already answered.

Brown’s order of operations follows directly: wire first, trade press second, website third. The sequencing runs against the instinct of most supply chain marketers, who default to redesigning the site.

“The LLMs don’t care about your website. They’re not going to your website,” she said. “The content on your website is important, but the order of operations is: send more press releases, get picked up by the trade media, and then make sure that you have pretty good content on your website.”

The reasoning is arithmetic and not aesthetics. No model is going to crawl 50,000 broker websites to find the one that answers a niche question in the fraction of a second it has to respond. It will reach for the wire and the trade desk that already did the work.

The consequence for an industry that has spent 10 years writing trade coverage off as legacy media is uncomfortable. “Trade press is more important than it was a year and a half ago,” Brown said.

The Index Is Where the Number Comes From

If the rule is to publish a number, the operational question becomes where the number comes from. Brown’s answer is an index, and she recommends one to nearly every company she talks to.

Two structures have worked. The first is mode-specific, and the discipline lies in picking a lane nobody owns.

An example she gave: Competing with DAT on macro rate data is a losing proposition, and the talent that built that advantage has since spread across the industry. Ken Adamo, DAT’s former chief of analytics and general manager of its shipper business, joined EASE Logistics as chief strategy officer in May. “He’s crushing it,” Brown said.

Competing on a mode that has not been claimed is a different proposition entirely.

ITS Logistics built its Port/Rail Ramp Freight Index into effective ownership of drayage commentary, to the point that Paul Brashier appears in the news whenever something breaks at a port. Brown points to heavy and light final mile as territory still sitting open.

The second structure is industry-specific, and the math is smaller than most marketers expect. One LeadCoverage client draws roughly 30% of its book from steel.

There are 123 steel manufacturers and distributors in America.

“The niche inside the niche is so small,” Brown said. The program reaches 1,200 to 1,400 people a month with an index on what is moving steel transportation rates: energy prices, fuel, disruption in the Middle East. One new customer a quarter clears the bar.

The origin story she returns to is Redwood Logistics, a client since 2020. Redwood had an internal cross-border newsletter, something built for internal use, not for outside eyes. It was a straightforward rundown of numbers and shipping data, nothing designed to be published.

Brown and her team saw an opportunity in it. They proposed repurposing that internal newsletter into something external: an index the industry could actually use. That idea became Redwood Mexico, which eventually brought CNBC to the company’s Laredo operation for a mini-documentary.

“The PR works if you are sharing regularly with the press a point of view with an economic perspective that matters to the shipper,” Brown said.

The Measurement Trap That Kills AI Citation Programs

The uncomfortable part is that the industry cannot yet prove any of this the way a CFO wants it proved.

Attribution tracking for citations is, by Brown’s own assessment, not yet reliable, and she is candid about the gap. The tooling is roughly six months old. Agencies can see citation volume and which model delivered it. What happens next is dark.

“We can tell you how many times you’ve been cited, but we can’t tell you what happens after that,” she said.

A buyer might click. A buyer might write the name down, or drop it into an analysis that goes to a boss. None of that is visible today.

Meanwhile, the numbers that are visible look wrong to anyone still grading on the old scoreboard. LeadCoverage’s search impressions rose 83% during the test. Google clicks fell, because AI answers were resolving buyer questions before a click could happen. Direct and brand traffic kept climbing, which is what it looks like when buyers find you somewhere other than search.

Read that dashboard with 2019 assumptions and it reads like failure. Brown expects that misreading to be the industry’s costliest mistake.

“The most common mistake we expect to see is companies grading these programs on clicks and shutting them down right as they start working,” she said. “Clicks are declining across the board. The measure that matters now is whether AI cites you when a buyer asks about your category. The source cited today is hard to unseat tomorrow, because these systems reward freshness and repetition. This is a position to claim before a competitor claims it.”