Writing for the AI That Finds You: the GEO Copywriting Guide

Berto rang me at the end of March. Going from memory: it was a Tuesday, I still had my coffee in hand, and outside the window Caserta was making up its mind whether to rain or not — that grey that wraps around the Reggia and leaves everything feeling a bit suspended. Berto isn't one for beating about the bush. He's got the hands of a man who's worked aluminium for thirty years and the healthy scepticism of someone who's watched too many consultants walk in with a tight tie and not much else to show for it.

"Umberto," he said, in that voice that always sounds a bit roughed up by the noise of the workshop, "I've done everything you told me. Blog kept up, keywords sorted, SEO in order. Three years of work. But when I ask ChatGPT who makes aluminium window frames in Santa Maria Capua Vetere... it can't find me."

There were ten seconds of silence. A proper Campanian silence — the kind that weighs as much as a philosophy treatise left out on a workbench.

"I don't exist?" he asked.

That question stuck in my head for the rest of the day, even as I was driving towards Naples with the smell of the sea drifting up from Via Caracciolo.

Berto absolutely exists. He has a website that ranks well on Google, a blog with forty carefully written articles, and reviews that would make a multinational jealous. The problem isn't the quality of his work. The problem is that his content was written for a reader who's disappearing — or who has at least changed habits quite radically. Berto had been writing for a user who googles, clicks and browses. But today, that user splits their time with a very different kind of interlocutor: an artificial intelligence that doesn't read the SERPs, doesn't compulsively click on the first three results, and doesn't bounce from page to page.

This artificial intelligence flicks through Berto's text like a university student at three in the morning before an exam: not looking for inspiration, but for a precise answer to extract, cite, and hand back to whoever asked the question. If that answer isn't served up on a silver platter — with the right structure and the right informational density — Berto is simply background noise to the AI. Invisible.

The same thing was happening to Bobby, in Salford. Bobby runs an HR consultancy and has that same lost expression when he uses Perplexity to search for a supplier and his name doesn't come up, despite years of investment in content marketing.

We're entering the era of generative search — technically known as GEO, Generative Engine Optimisation — and the rules of the game changed while we were still getting our heads round the old ones. It's no longer enough to "write well". You need to write to be found by someone who doesn't have time to read everything you've written, but has the power to cite you.

In this article I won't be covering GEO as a marketing strategy — for that, there's my dedicated piece on local GEO and territorial marketing. Today I want to get to the heart of the matter: the writing itself. How do you write a paragraph so that ChatGPT decides you're the authority worth citing? How do you structure information so that Google AI Overviews puts you front and centre? This isn't about abstract algorithms — it's the same pragmatism Berto uses when he needs to cut an aluminium profile. If the dimensions aren't right, the piece won't fit.

The reader you can't see

When we write, we always have a mental image of who we're talking to. Usually it's a human being — perhaps a bit distracted — scrolling through their phone while waiting for a train at Piazza Garibaldi or nursing a pint in a Shoreditch pub. But today we have to reckon with the existence of a shadow reader: one with no emotions, no appreciation for poetic nuance unless it serves a purpose, and an insatiable appetite for data and logical structures. This reader is the generative search engine. And it's the first reader of your text.

The numbers that make this urgent

This isn't just my romantic impression. The data from 2025 and 2026 tells us clearly that the ground is shifting beneath traditional copywriting:

• ChatGPT handles around 2.5 billion prompts per day — a public estimate from July 2025 — and according to OpenAI's March 2026 update, use of the integrated search function has nearly tripled over the past year
• Traffic generated via AI channels grew sharply throughout 2025: the Previsible AI Traffic Report, based on analysis of nearly 2 million LLM sessions, documents strong growth across all generative search channels — with some segments recording year-on-year increases running into three figures
• Several analyses from 2026 — including estimates from Superlines and Position Digital — indicate that over 90% of AI search sessions end without the user clicking on any website; the precise figure varies between sources and hasn't yet been consolidated by a single primary study, but the direction is consistent: zero-click search in AI search is structurally higher than in traditional Google
• Those who arrive at your site via an AI citation convert on average 4.4 times better than traditional organic Google traffic (Semrush AI Search Study, 2025/2026) — the scope of the study isn't universal, but the qualitative signal is clear: anyone arriving via AI is already pre-filtered and far further along in the decision-making process

That last figure is the most important one for Berto — and for Bobby. The AI acts as a filter: whoever cites you hands you a client who's already half-convinced. Volume goes down, quality of leads goes up dramatically.

How RAG (Retrieval-Augmented Generation) actually works

Before understanding how to write, you need to understand who is reading. The AI doesn't know everything. When a user asks a question, the engine uses a process called RAG — Retrieval-Augmented Generation: a very fast assistant that dashes to the library, picks up five books, reads the relevant bits, and produces a summary. Here's how it works in detail:

1. Query fan-out — the question multiplies.

The AI doesn't pass the entire query to the search engine. It breaks it down into smaller sub-questions and searches for them separately. "Best plumber in Portici" becomes three distinct searches: "plumbers Portici", "plumber reviews Naples", "plumbing installation costs Campania". If your text doesn't answer any of these fragments explicitly, you don't exist in any of the three searches.

2. Document retrieval — the detail that changes everything.

This is the critical point that almost nobody considers. The system doesn't read your entire page: it extracts paragraphs. An analysis by Kevin Indig (Growth Memo, February 2026) covering over 1 million queries and 1,012 confirmed citations from ChatGPT shows a clear distribution of where LLMs go to pull information:

The practical takeaway is brutal: putting key information at the bottom of a page reduces the chances of being cited by 2.5 times compared to placing it at the top. If your answer is buried in paragraph eight, after three warm-up paragraphs of historical context, the AI has already moved on.

3. Synthesis — who gets chosen.

The AI combines information from three to seven different sources into a coherent response. Who gets cited? Not necessarily whoever ranks first on Google, but whoever has the clearest, most structured, and most verifiable answer.

Think of Bobby's deputy in his Hackney office: he hasn't got time for a forty-page report. He wants the three-line summary sitting on top of his desk. If it's not there, he moves on to the next supplier without looking back. The AI works in exactly the same way. Understanding this mechanism changes everything: we're not writing "articles" — we're building extractable information modules. Every section needs to stand on its own and be complete enough to be lifted out and used as a building block in a generated response.

ChatGPT, Perplexity and Gemini aren't the same thing

This is the point that almost no guide — Italian or otherwise — has yet addressed properly: the three major AI engines cite sources in completely different ways. Treating them as a single entity means optimising poorly for all three.

According to an analysis by Yext covering 6.8 million citations (October 2025):

• Gemini primarily trusts the brand's own website: 52.15% of its citations come directly from the company's domain. Investing in the quality of your own site is the main lever to pull.
• ChatGPT trusts what the web agrees on: it prefers third-party sources, Wikipedia, Reddit, forum mentions. A craftsman from Caserta who appears on Wikipedia or gets cited in a Reddit thread about made-in-Campania products gains a direct advantage.
• Perplexity trusts niche experts and sector-specific reviews. For a restaurant owner in Naples: TripAdvisor, food guides, mentions in trade publications. For an HR consultant in Salford: Glassdoor, articles in specialist HR outlets.

Practical tip: a workshop owner in Portici who wants to show up on all three works on distinct levels. On their own site, for Gemini. On Wikipedia and Reddit, for ChatGPT. On sector directories and reviews, for Perplexity.

SEO and GEO: writing for two readers with different goals

A lot of people ask me whether old-school SEO is dead. The answer is no — but it's become the foundation on which to build something more complex. If classic SEO was a sprint to the top of a hill, GEO is a competition on usefulness — the prize being an invitation to the table. SEO focuses on where you rank; GEO focuses on what you say and how you say it.

We don't have to choose between one approach and the other. Sites with solid traditional SEO rankings start with an advantage because AI crawlers visit them more frequently. However, that advantage is eroding fast. According to Ahrefs, between mid-2025 and early 2026, the percentage of Google AI Overviews citations drawn from the top ten organic positions dropped from 76% to around 38%. A good ranking is no longer a safe harbour — it's just a ticket to be in the running. To win, you need substance.

The Portici case: the difference between searching and asking

Imagine an estate agency in Portici, ranking first on Google for "flats for rent in Portici". But when someone asks Perplexity "what are the best areas to live near Naples without spending a fortune, with a train station nearby?", the AI doesn't just show property listings. It looks for a text that explains what life in Portici is actually like — the rail connections, the average price per square metre compared to Naples city centre, how safe the neighbourhoods are. If the agency only has technical property sheets, it'll never get cited. If it has an article containing a data point like "the average price in Portici is 20% lower than in the Vomero district, with the Circumvesuviana connecting to the city centre in 18 minutes", it becomes the authority the AI cites.

Writing for generative search means stopping thinking about keywords and starting to think about answers. For a deeper look at the relationship between persuasive writing and algorithmic visibility, have a read of this article on balancing SEO and persuasion — the tension between the two approaches is subtler than it looks.

The 7 rules of GEO-friendly writing

Writing for the era of generative search isn't some dark art. It's a discipline built on precision and respect for time — including an algorithm's time. Here are the seven rules I've developed working between Caserta and London, testing them in the field with real results.

Rule 1: the answer first. Context after.

AI systems like ChatGPT and Google AI Overviews strongly favour content that front-loads the answer. This means reversing your narrative instinct. We're used to setting the scene, providing context, telling the family backstory before getting to the point. Forget all that.

Research confirms that AI systems prefer pages that answer the query within the first 40–60 words of the relevant section, before expanding with context and detail.

Before/after example — craftsman from Caserta:

Before (classic SEO writing):

"In the competitive landscape of Caserta's ceramic production — which boasts a centuries-old tradition rooted in local craftsmanship and the skilled hands of our forebears — many today find themselves facing new challenges... the ceramics produced in Caserta are distinguished by..."

After (GEO writing):

"Artisan ceramics in Caserta are produced primarily in Santa Maria Capua Vetere and Capua, using traditional techniques passed down through generations. Local producers offer tiles, decorative objects and bespoke commissions. Average prices start from €35 per piece for batch production."

The second text is extractable. The AI can use it directly. The first is just atmosphere. Don't be afraid to be direct: the human reader will thank you for it and the AI will reward you.

Rule 2: modular sections, one question per section

Every section of the article must answer a single specific question — the kind the AI might use as a sub-query in its fan-out analysis. If the article is one unbroken block of 3,000 words, it's hard for the crawler to digest. Split into blocks of 75–150 words, each with a clear heading, it becomes a goldmine of citations.

This modular logic shares its DNA with the principle of "show, don't tell": both force you to make information visible rather than submerging it in context. If you'd like to explore how to apply this principle to the visual structure of your content as well, you'll find everything in the article The Art of "Show, Don't Tell" — it's the technical foundation from which GEO writing also flows.

Rule 3: one data point every 150–200 words, always cited

According to the Princeton research that coined the term GEO (2024), including statistics with a verifiable source increases visibility in AI results by 30–40%. The data point shouldn't be decorative — it should be the load-bearing structure of the sentence.

Wrong:

"Many Italian SMEs are investing in digital."

Right:

"According to the Confindustria Digitale Report 2025, 43% of manufacturing SMEs in Southern Italy increased their digital budget over the past year, with an average increase of 18%."

This gives the AI a "proof of authority" it can use to justify its response to the user.

Rule 4: H2/H3 headings as real questions

The AI looks for semantic matches. If a section heading is a genuine question — "How does the renovation process work for a flat in Naples?" — you dramatically increase the likelihood that your paragraph gets extracted as a direct answer. Avoid clever but opaque headings like "The Breath of the Walls". The AI has no imagination — only logic.

Wrong: H3: Our kitchen

Right: H3: What dishes do you serve and how do you handle allergens?

Rule 5: precise, consistent terminology. No creative synonyms.

Classic creative writing teaches us to vary our terms to keep things interesting. GEO writing wants consistency. The semantic matching of AI engines recognises the concept, but unstable terminology weakens the signal.

Before/after example — agency from Santa Maria Capua Vetere:

Before ("stylistic" variation):

"Our communications agency is the creative studio that makes the difference. Our team — a true boutique of Caserta marketing — works on every project..."

After (terminological consistency):

"Our communications agency in Santa Maria Capua Vetere works on branding, digital communications and advertising campaigns for SMEs across Campania. The agency has been operating since 2008."

In the second text, "communications agency" appears twice. That's not repetition — it's a signal. The AI knows exactly what this business is and where it operates.

Rule 6: cite authoritative external sources in the text

It sounds counterintuitive — why send readers elsewhere? But citing authoritative third-party sources (universities, research institutions, official publications) places you within a worthwhile conversation. AI systems perceive those who cite credible sources as more trustworthy, much like in academic research. Every long-form article should include at least 2–3 external links to primary sources.

Rule 7: update regularly. The timestamp is a sign of life.

50% of content cited by AI search is less than 13 weeks old (Amsive, 2026). Pages updated within the last 60 days are cited 1.9 times more often than pages that haven't been touched (BrightEdge, 2026). Adding a "What's changed in 2026" section to an old article and updating the modified date is one of the quickest ways to become relevant again. The AI hates stale content just as much as Bobby hates a warm pint.

Who's doing it well and who's invisible: real examples

To properly understand how GEO works, we need to look at who's winning this quiet battle — and who's losing ground despite million-pound budgets. It's not a question of company size, but of editorial intelligence.

Positive example 1: Tally (SaaS form tool)

Tally is an online form-building tool that made a radical editorial choice: all its documentation and blog content follows the principles of absolute clarity and immediate answers. If you search "how to create a payment form with Tally", you find a page where the answer is in the first line, followed by precise technical data. Zero atmosphere, maximum extractability.

Result: by 2026, ChatGPT had become Tally's number one user acquisition channel, overtaking organic Google and social media. AI traffic accounts for around 10% of new monthly sign-ups (source: Tally case study, Search Engine Land, 2026). They understood that today's user doesn't want to browse the site — they want the AI to tell them how to do things using Tally.

Positive example 2: Vercel (cloud infrastructure)

Vercel built an editorial strategy around modular technical content: every section answers a specific question from developers, headings are framed as real queries, data is always cited with a source. ChatGPT and Perplexity cite them constantly. Referrals from ChatGPT went from 1% to 10% of new sign-ups within six months in 2026. It's not just brand authority — it's text structure.

Negative example: the institutional copywriting pattern (documented typical case)

I regularly work with mid-range brands — Italian and British — that all follow the same pattern: excellent SEO ranking, zero AI citations. Italian institutional copywriting tends to build suspense before delivering the answer. Phrases like "our spring-summer collection celebrates timeless elegance and the joy of Mediterranean living" are gorgeous in a printed catalogue. For an AI, they're background noise.

In the pattern I observe most frequently — verified across several fashion and retail brands in the analyses I carry out between Naples and London — the AI systematically ignores pages with this structure and instead cites micro-influencer blogs that have written comparative guides with concrete technical data: fabrics, precise prices, verifiable links. The AI has no aesthetic taste. It only has the ability to extract answers. And it finds those answers wherever someone has taken the trouble to write them.

The Aranzulla case: when you write well for the AI but the AI doesn't return the favour

There's an Italian case worth examining carefully, because it's more complex — and more instructive — than any straightforwardly positive or negative example.

Salvatore Aranzulla built one of Italy's most visited tech explainer sites by writing in exactly the way GEO demands: immediate answer in the first paragraph, modular structure, headings framed as real user questions, consistent terminology, no warm-up paragraphs. "Write like Aranzulla" was for years the unwritten rule of Italian SEO copywriting. And he was right.

The problem arrived in October 2025, when in an interview with Fanpage.it he reported a 25% drop in traffic to his site, which he attributed to Google AI Overview. His diagnosis was surgical: "For years there was an implicit agreement between Google and those who publish content online. We provided free content and Google sent traffic. That mechanism no longer holds."

The paradox is precise: aranzulla.it writes so well — with such clear structure and such extractable answers — that the AI lifts his information, synthesises it, and delivers it to the user without that user needing to click through to the site. The content gets used, but the traffic doesn't arrive. It's the exact opposite of Berto's problem: Berto wasn't being found because he was writing poorly for the AI. Aranzulla is found by the AI constantly, but his business model — built on advertising revenue and page view volumes — suffers precisely because the AI uses him too well.

The lesson for anyone reading this article is twofold. In the short term: optimising for GEO brings citations, and citations bring high-quality traffic (that 4.4x conversion rate we mentioned earlier). In the long term: anyone whose business is built entirely on display advertising tied to clicks needs to rethink the model, not just the writing. Those are two distinct problems. The first — how to write — is what this article deals with.

The Wikipedia case: the perfect (unintentional) model

Why do AI systems cite Wikipedia almost compulsively? Not because it's been optimised for SEO. Because it follows — entirely unintentionally — every rule of GEO: immediate answer in the first paragraph, modular structure with clear H2/H3s, data cited with precise sources, consistent terminology, frequent updates.

The practical takeaway for Berto: write as if you're writing for Wikipedia. Not in tone — which can stay warm and unmistakably Campanian — but in structure. Clarity is the only currency the AI accepts without haggling.

The technical layer that amplifies your writing

You can write the best copy in the world, but if the "house" hosting it has the doors bolted shut, the AI will never get in. There are four technical elements to check immediately — and at least one of them is almost certainly a problem on your site.

Robots.txt and Cloudflare: don't block AI crawlers without realising it

Lots of sites are blocking AI crawlers without even knowing it. From mid-2024 onwards, Cloudflare changed the default configuration for many plans, adding rules that automatically block AI bots for "content protection". The problem: if you block the bots, you disappear from the answers.

How to check: in the Cloudflare dashboard, look for the Security > Bots section or the dedicated AI metrics page (the exact name varies with platform updates — search for "AI Scraper" or "Bot Fight Mode"). Make sure the user agents ChatGPT-User, PerplexityBot and Google-Extended aren't being blocked. At the same time, open your server logs and look for these user agents: if they never appear, the workshop door is locked. No writing, however perfect, can get through a wall.

Schema markup: the code that speaks directly to the AI

Structured markup in JSON-LD format tells the AI what the content is, not just what it's about. The three most useful types for local SMEs:

FAQPage schema — for all question-and-answer sections. Increases AI citations by 44% (BrightEdge, 2025). Implemented by adding a JSON-LD block in the <head> of the page, or — if you're using WordPress — with plugins like RankMath or Yoast SEO that generate it automatically. Each question and answer in the markup corresponds to an H3 and the paragraph that follows it.

LocalBusiness schema — essential for anyone with a physical premises in Naples, Caserta, Portici or Santa Maria Capua Vetere. Includes name, full address, opening hours, geographical area served, phone number. This schema tells Perplexity — which favours local directories — that you're "the right one" for that area. Without it, the AI can't place you geographically with any precision.

Article schema — for blog articles and guides. Signals the author (name, profile), the publication date, and — crucially — the date of the last update (dateModified). This timestamp is one of the most direct freshness signals you can send to AI crawlers.

The llms.txt file: introducing yourself to the AI by name

A new standard is emerging — not yet universal, but growing fast: the llms.txt file in the root of the site. It functions like a direct business card for language models. It contains a concise description of the site, an indication of the content most relevant for citation, and links to the main pages in Markdown format. Here's a minimal example:

# Site name — Concise description in 2 lines

## Main content
- [Services page](https://yoursite.co.uk/services): brief description
- [Main guide](https://yoursite.co.uk/guide): brief description

## About us
Communications agency founded in 2008, based in Santa Maria Capua Vetere (CE).
Specialising in: branding, digital marketing, Campanian SMEs 

It's not compulsory, but it's a signal of openness that language models appreciate — and that almost nobody in Italy has yet implemented.

Server-side rendering: text hidden in JavaScript doesn't get read

AI crawlers, unlike browsers, don't execute JavaScript. If parts of the site only appear after a user interaction — hidden tabs, lazy-loaded content, animations that reveal text — the AI sees that page as partially empty. All relevant content must be present in the static HTML returned by the server, without needing to be "activated". If you're using a frontend framework like React or Vue without SSR (Server-Side Rendering), now's the time to have a serious conversation with your developer.

Three before/after rewrites

To give you something concrete and immediately actionable, here are three examples of the transformation from traditional writing to GEO-friendly writing. These are cases I come across every week, travelling between Campania and England.

Case 1: law firm in Naples (Piazza Municipio)

Target query: "Employment law solicitor in Naples"

Before (institutional):

"Studio Legale Esposito e Associati has operated in the heart of Naples since 1998, offering qualified legal advice across multiple areas of law. Our team of professionals guarantees personalised, competent assistance to individuals, companies and institutions, with an approach aimed at resolving disputes..."

After (GEO-friendly):

"Studio Esposito e Associati specialises in employment law in Naples, with offices in Piazza Municipio. It handles redundancies, contractual disputes and workplace harassment cases for employees and companies across Campania. Free initial telephone consultation. Response within 24 hours. Since 1998 we have handled over 1,200 employment cases before the Naples courts."

Case 2: artisan pasta producer, Gragnano

Target query: "Buy artisan pasta from Gragnano online"

Before (narrative):

"Our pasta tells the story of Gragnano, the pasta capital par excellence, where the wind from the Gulf of Naples would dry the sheets along historic streets filled with the smell of the sea and the Campanian sun. Every shape is the result of careful craftsmanship..."

After (GEO-friendly):

"Artisan pasta from Gragnano IGP available online with 48-hour delivery. Shapes: spaghetti, paccheri, rigatoni, fusilli. Made with 100% Italian durum wheat, bronze-die extruded and dried at 40°C for 36 hours. From €4.80 per 500g. Slow drying preserves the wheat's protein content."

Case 3: HR consultant, Salford (Manchester)

Target query: "HR consultant for small businesses Manchester"

Before (business jargon):

"At Northfield HR Solutions, we believe that people are the heart of every business. Our tailored approach combines deep expertise with genuine care for your team's wellbeing. We work with SMEs across Greater Manchester to create workplaces where people thrive..."

After (GEO-friendly):

"Northfield HR Solutions provides HR consulting for small businesses in Manchester and Salford. Services: employment contracts, disciplinary procedures, TUPE transfers and redundancy processes. Fixed-fee packages from £350/month. Specialising in companies with 5–50 employees across the North West."

Notice the difference? Fewer adjectives, more nouns. Fewer promises, more data. Less fog, more light. The tone can stay the same — warm, human, authentic — but the structure has to change.

How to measure whether it's working

A lot of people say to me: "Umberto, how am I supposed to know if the AI is finding me? Google Search Console doesn't tell you much yet." That's true — we're in something of a no-man's-land in terms of established tools. But there are three approaches that Berto and Bobby use with success — from the simplest to the most structured.

Monthly manual test (free).

Open ChatGPT, Perplexity and Google AI Overviews in a private browsing session. Type in the ten questions your typical customer would ask. See who gets cited. If it's not you, analyse who is there and how they write. It's the most useful competitive intelligence of the century.

Google Analytics 4 — AI referral segment.

In GA4, create a custom segment that filters sessions coming from AI domains: chat.openai.com, perplexity.ai, bard.google.com, bing.com/chat. It's not a complete picture, but it tells you whether AI traffic is growing month on month. If the curve is going up, your writing is working.

Dedicated tools.

If you've got the budget, platforms like OtterlyAI, Peec AI or AthenaHQ track citations across six to eight AI platforms simultaneously. They give you your Share of Model: how frequently your brand appears in responses to relevant queries. Prices are accessible even for individual practitioners — from around €50–80 per month for entry-level plans. The goal isn't 100%. It's to grow every month.

The three writing mistakes that make you invisible

To close, here are the three "cardinal sins" I see repeated obsessively in Italian and British company blogs. These are exactly the mistakes that had kept Berto invisible for months.

Mistake 1 — The warm-up paragraph.

The habit of opening every section with a philosophical or historical reflection. "In modern society, aluminium plays a crucial role...". The AI reads the first three lines. If the answer isn't there, it moves on. Cut the warm-up and get straight into it.

Mistake 2 — Terminological variation to sound literary.

Using "office", "commercial premises", "operational hub" and "studio" in the same text confuses semantic matching. Terminological consistency is your best ally: choose the name your clients use and don't change it for aesthetic reasons.

Mistake 3 — Data without a source.

Writing "according to recent studies" without specifying which ones is like saying "a mate told me". The AI looks for verifiable sources to protect its own credibility. "According to the ISTAT 2024 Report on Business Structure..." cites someone real. Same information, completely different value.

Berto and Bobby, twelve months on

Twelve months ago, Berto didn't exist for ChatGPT. He was a digital ghost with a beautiful website that no AI was reading. Today, if you ask Perplexity who makes aluminium window frames in Santa Maria Capua Vetere, his name comes up — not always in first place, but often enough to have brought him three new quotes in the past month from people who'd never have found him through traditional SEO alone.

Berto didn't rebuild his site from scratch. He rewrote the key sections following the rule of the immediate answer. He added data on the average lifespan of his window frames, citing laboratory test results. He added LocalBusiness schema for his Caserta premises. He checked the robots.txt and discovered that Cloudflare had been blocking AI crawlers for months without anyone noticing.

Bobby, in Salford, did the same with his HR consultancy. Now 12% of his new enquiries come from generative search — a channel that six months ago was a black hole in his analytics.

Take your most visited article. Rewrite the first paragraph so the answer is in the first two lines. Add an FAQPage schema. Check the robots.txt. That's three hours of work — less than a morning. Don't wait for your competitor in Santa Maria Capua Vetere or Salford to do it before you.

The window is still open. Most of your competitors are still writing "360-degree solutions" and "industry leaders" — phrases the AI dismisses with a quiet shudder of digital boredom.

Now you know where to look. The next time you write an article, ask yourself: if ChatGPT only read the first two lines of this section, would it know what I do and what I'm worth? If the answer's no, go back and write it again.

I'll be here — probably somewhere between a coffee in Caserta and a pint in London — watching how the world changes, one line at a time.