A rate change you announce at nine in the morning can be inside an AI answer before your afternoon meeting ends. Speed is now part of visibility, and structure is what makes speed possible.
Eight hours.
That is how long it took, on average, for a press release to get its first AI citation in Notified’s June 2026 study of 8,000 GlobeNewswire-distributed releases. Most citations landed within the first 24 hours. Some compliance reviews I have sat through took longer than that.
Here is the stat worth keeping: according to Notified’s June 2026 analysis of 8,000 press releases, the average time from publication to first AI citation was eight hours, and 99.3% of those releases were eventually cited by ChatGPT or Claude.
I want to sit with that second number for a second. Ninety-nine point three percent. AI systems are not being picky about press releases. They are ingesting nearly all of them. The competition is not whether your announcement gets read by the machines. It is whether the machines can pull anything useful out of it.
The part that should get your attention
Notified and PRWeek followed up with a case-study report on July 22, 2026, built on what they call the SOAR framework: structure, originality, authority, recency. The framework itself comes from an analysis of more than 250,000 press releases and nearly a quarter of a billion AI citations.
One example from that work stuck with me. An earnings release from American Battery Technology Company, a company most people could not pick out of a lineup, generated more than 1,600 AI citations in 30 days. It outperformed a comparable peer and a nationally recognized retail brand.
I have been saying a version of this for a year: brand equity does not predict AI visibility. A famous name gets you nothing if your content is unreadable to the systems doing the reading. A small institution with clean, structured announcements can out-cite a household name. Now there is press release data saying the same thing.
The report also includes a case study from Ally Financial, so this is not a retail-brand phenomenon that skips our industry. Financial institutions are already playing this game. Most just have not looked at the scoreboard.
What this means if you run marketing at a bank or credit union
Think about what you announce in a normal quarter. A CD special. A rate change on a money market. A new checking product. A branch opening.
Every one of those is exactly the kind of content AI systems love to cite, because it is specific, dated, and verifiable. A rate is a number with a name attached. That is the easiest thing in the world for a machine to extract and repeat.
And the eight-hour window changes the math on speed. When your competitor across town publishes their new CD rate in a structured announcement, that rate can be in ChatGPT’s answer to “best CD rates near me” by dinner. Yours, sitting in a PDF three clicks deep on your site, is invisible. Not penalized. Just never seen.
The same-day loop, in practice
You do not need a wire service subscription to act on this. You need a repeatable habit for announcements:
Put the number in the first sentence. Rate, fee, date, dollar amount. Do not make anyone, human or machine, hunt for it.
Publish on your own domain as a real page. HTML, not a PDF. A newsroom or blog URL that crawlers can reach.
Structure it. Headings, Article schema, your institution’s full name spelled the same way every time. Name the product like you want it quoted.
Refresh it when things change. Recency is one of the four signals in Notified’s framework. An announcement from 2023 about rates that no longer exist is worse than nothing.
None of this is exotic. It is the same structure-first work I keep coming back to, because in the audit work we do at Atlas Instinct, structure is where financial institutions lose the most ground and where fixes pay off fastest. The firm’s breakdown of the five signals AI systems weigh before recommending an institution is the natural companion read.
The caveat, because there always is one
Notified sells press release distribution. Their study says press releases work. Keep that in mind, the same way you would with any vendor research, and I have framed every number here with its source and date on purpose. But the direction of the finding matches what I see in audit data that has nothing to do with press releases: structured, specific, dated content gets found. Vague content does not.
The eight-hour clock is real either way. The only question is whether anything you published this quarter is legible enough to be on it.
If you are curious where your institution stands, the 90-second self-check at kevinfarley.org/ai-visibility-score is a decent place to start. And if you have a rate announcement going out this week, try the loop above on it. Same-day is not a stretch goal anymore.