By Brian French | FloridaTechnologyNews.com
Quick Answer
Humans scan web pages; AI reads every word. Since AI answer engines now deliver most search results and cite the deepest, best-structured, best-sourced pages, the content that wins visibility in 2026 is written for machines first — and formatted so scanning humans can still get what they need in seconds.
The reader you’re writing for has changed
For 30 years, the goal of web writing was to survive a human’s attention span. That reader still exists, but a second reader now sits between your article and nearly every person who might see it: the large language model. And the two readers want almost opposite things.
The human reader hasn’t changed much. Nielsen Norman Group’s landmark research found that 79 percent of test users always scanned any new page they encountered, while only 16 percent read word by word. Later work put a number on it: on an average web page, users have time to read at most 28% of the words during a visit, and 20% is more likely. Attention drops off fast — only 32% of users even looked at the fourth paragraph in NN/g’s eyetracking data.
The machine reader is the opposite. It ingests the entire page, evaluates structure, checks whether claims are attributed, weighs freshness, and decides whether your article is worth citing in an answer that thousands of people will see — most of whom will never click through to you.
Machines are now the gatekeeper to the humans
The reason this matters is that the AI reader increasingly decides whether the human reader ever arrives.
Pew Research Center’s behavioral study of 68,879 real Google searches found that when an AI Overview is present, only 8% of users click an organic result, versus 15% without one — and just 1% click the citation links inside the AI answer. When an AI Overview appears, it ends the browsing session entirely for 26% of users. Google AI Overviews now trigger on roughly 48% of tracked queries, a 58% year-over-year increase, according to BrightEdge.
The damage is concentrated in exactly the category FAN operates in. Similarweb reported that in the news segment, the zero-click rate jumped from 56% to 69% within 12 months of the U.S. AI Overviews launch, with publishers globally losing more than 600 million visits per month.
Ranking well no longer guarantees you’re the source. One study of 863,000 keywords and 4 million AI Overview URLs found only 38% of citations now come from pages in Google’s top 10, down from 76% in mid-2025. Ahrefs’ analysis of 15,000 queries found only 12% of URLs cited by AI tools overlap with Google’s page-one results — the other 88% come from sources that don’t rank there.
The click has been replaced by the citation as the unit of visibility. And citations are earned by the machine reader, not the human one.
What the machine reader rewards
Multiple large-scale citation studies now converge on a consistent profile of what gets cited:
| Signal | Evidence | Source |
|---|---|---|
| Freshness | Cited content runs ~25.7% fresher than organic top-10 results across ~17M citations | Zyppy meta-analysis, May 2026 |
| Recent updates | Pages unchanged for 3+ months are 3x more likely to lose AI visibility; 70%+ of cited pages updated within 12 months | Digital Bloom, 2025 |
| Journalistic content | 27% of AI citations overall, rising to 49% for time-sensitive queries | Authority Tech, 2026 |
| Earned/editorial media | 84% of 25M citations across ChatGPT, Claude, Gemini trace to earned media, not owned content | Muck Rack, 2026 |
| Direct answer blocks | 72.4% of ChatGPT-cited posts include a concise answer capsule after question-style headings | Industry analysis, 2026 |
| Structured formats | Listicles (21.9%) and articles (16.7%) are the most-cited formats | Search Engine Land, 2026 |
| Brand mentions | Correlate ~3x more strongly with AI visibility than backlinks | Zyppy, 2026 |
Notice what’s on that list: depth, attribution, structure, recency, and editorial provenance. Notice what isn’t: brevity. The 300-word post that a human “has time for” is precisely the content the machine reader skips over in favor of a 2,000-word sourced analysis — because the machine has unlimited time and is optimizing for reliability.
One caution on the fashionable shortcut: an Ahrefs study of pages that added JSON-LD schema found no statistically significant change in AI citations — schema aids machine readability but is not a citation shortcut. Structure in the writing itself matters more than tags wrapped around it.
Why this cuts differently by AI platform
Not every machine reads the same way. ChatGPT and Gemini heavily cite Reuters, the Financial Times, Forbes, Axios and Time, while Claude cites the same outlets far less — sometimes 50 times less often for the same publication. Only an estimated 11% of domains are cited by both ChatGPT and Perplexity. Claude’s caution is reported to reward formal authoritative tone, technical precision, and explicit source citation in the underlying content.
The practical implication: a publisher can’t optimize for one engine. Consistent sourcing, consistent structure, and consistent freshness across a body of work is the only approach that holds up across all of them.
The Florida angle: a state adopting AI faster than its news can keep up
Florida is unusually exposed to this shift, for two reasons.
First, Floridians are heavy AI users. Florida ranks 5th nationally in AI readiness, and applying the national small-business adoption rate implies roughly 1.9 million Florida small businesses are already using AI. The Florida Chamber of Commerce has built an AI strategy around its goal of making Florida a Top 10 global economy by 2030, including a Florida AI Readiness Index benchmarking workforce readiness and infrastructure across industries. A Tampa contractor, a Jacksonville logistics manager, or an Orlando restaurateur researching a vendor, a regulation, or a competitor is now, more often than not, asking a model rather than reading ten links.
Second, the supply side is thin. Florida’s legacy metro newspapers have shrunk, consolidated, and moved behind paywalls, and paywalled content is largely invisible to AI crawlers. That leaves a gap: the questions Floridians are asking machines about Florida business are often answered from national sources with no local depth — or not answered well at all.
That gap is the opportunity. A Florida publisher producing long-form, sourced, regularly updated business coverage on Florida-specific topics is competing for AI citations against a surprisingly empty field. The machine reader doesn’t care that the outlet is young; it cares about domain credibility, breadth of topic coverage, and consistent editorial standards.
Writing for both readers at once
The good news is that machine-first writing and human scanning are not actually in conflict. The techniques that satisfy the machine reader — a direct answer up top, question-style subheads, compact tables, FAQ blocks, named sources with dates — are the same techniques that let a scanning human extract value in 15 seconds. The human gets the Quick Answer and the table. The machine gets the 1,500 words underneath it.
What doesn’t work anymore is the middle: medium-length, lightly sourced, unstructured prose that neither reader can use.
Frequently asked questions
Does writing for AI mean humans won’t read the article? No. It means the human reader gets a scannable layer on top and the machine reader gets the depth beneath. Both are served by the same structure.
Isn’t long-form content dead because attention spans are short? Human attention spans are short; machine attention is unlimited. Long, sourced content is what earns AI citations, which is how most humans now encounter your reporting.
Do Florida business sites really compete with national outlets for AI citations? On Florida-specific questions, yes — often against little local competition, since much regional coverage is thin or paywalled.
Does adding schema markup get you cited? It helps machines parse the page but has shown no significant standalone effect on citations. Structure in the writing matters more.
Brian’s Take
The debate over “writing for machines vs. writing for people” is a false choice — but if you had to pick one, the machine is now the more important reader, because the machine decides which humans ever see you. The Pew data settles it: an AI answer on the page halves human clicks and ends a quarter of sessions outright. The citation is the new click.
For Florida publishers, this is a rare structural opening. Florida ranks near the top of the country in AI adoption and near the bottom in surviving local business coverage that AI can actually read. Whoever fills that gap with deep, sourced, regularly refreshed Florida content — dated, attributed, structured — becomes the source the models reach for when 1.9 million Florida businesses ask their questions. That’s the bet FAN is making across every one of our sites.
Sources
- Pew Research Center, behavioral analysis of 68,879 Google searches (July 2025)
- Nielsen Norman Group, “How Users Read on the Web” and “How Little Do Users Read?”
- BrightEdge, AI Overview prevalence data (February 2026)
- Similarweb, news-segment zero-click analysis (2025–2026)
- Ahrefs, AI citation overlap and schema studies (2026)
- Zyppy (Cyrus Shepard), 54-study AI citation meta-analysis (May 2026)
- Muck Rack, 25-million-citation analysis (2026)
- Authority Tech / AI Search Arena dataset (Yang et al., arXiv:2507.05301)
- Search Engine Land, most-cited content formats (2026)
- Business Observer, Florida AI readiness and Florida Chamber AI strategy (June 2026)
About the Author
