Your Intro Has Two Readers Now. Most People Are Writing for One. | Soaring Above Search #3

I spent part of this week rewriting intro paragraphs.

Not because the articles were bad. The topical coverage was solid, the research was there. But when I ran them through AI search, ChatGPT was skipping the intro entirely and extracting from the first H2 instead.

The intros were written for humans. Not for the LLM deciding whether your page earns a citation.

In 2026, every intro paragraph has two jobs.

The Two Readers Problem

The first reader is the human who clicked your link. They're deciding in the first five seconds whether you're going to answer their question or waste their time.

The second reader is the AI model deciding whether your page is worth citing. It's scanning your intro for a self-contained factual claim that directly answers the query it received. If it doesn't find one, it moves to the first H2.

Most intros handle the human well. They have a hook, some subtext, maybe a proof element. What they don't have is a self-contained answer in sentence one that an LLM can extract and use verbatim.

Matt Diggity's intro checklist covers the human side comprehensively. The hook, the subtext, the proof element, the empathy, the transition. Run that checklist and your bounce rate improves. But your AI citation rate doesn't move until you add the second layer.

What LLMs Actually Do at the Intro Level

When a RAG pipeline retrieves your page, it chunks the content and scores each chunk for relevance to the query. The intro gets processed first. If it's evasive — building to an answer rather than opening with one — the model gives it a low relevance score and moves on.

Hostinger ran a three-month experiment across 100 pages, using answer-first introductions as one of five structural changes. Citation share jumped from 3.1% to 4.7%, a 52% increase. Structure drove the gain, not additional authority or links.

This isn't a soft best practice. It's a chunking and relevance score problem. The fix is mechanical.

BLUF: The Missing Layer

BLUF stands for Bottom Line Up Front. It's how military briefings, good journalism, and technical documentation are written. State the conclusion first, then explain it.

In 2026 SEO terms: sentence one of your intro should be extractable on its own and still answer the query. Everything that follows — the hook, the context, the empathy, the subtext — supports that opening claim.

The formula: BLUF sentence, then human hook, then context, then transition.

Most intros today are structured backwards. Hook, context, answer buried somewhere in paragraph three.

What This Looks Like in Practice

Three real examples. The before version is how each article was structured. The after version is what we changed it to.

Example 1: AI Overviews explainer

Before: "Have you ever clicked on a Google result only to find it didn't actually answer your question? You're not alone. In 2026, Google has fundamentally changed how search results work, and understanding these changes is critical for any business trying to get found online."

After: "AI Overviews now appear on 99.9% of informational search queries. For content that earns a citation, that means a 35% CTR boost. For content that doesn't, it means organic traffic diluted by an answer that never sent a click. Here's how to end up on the right side of that split."

The after version gives the LLM something to extract on line one. The before version makes it wait three sentences for anything useful.

Example 2: ChatGPT data sources article

Before: "Curious where ChatGPT's knowledge comes from? You're not alone. This article breaks down the types of data used to train ChatGPT, how current its information is, and why it all matters for output quality. Whether you're an SEO, researcher, or AI enthusiast, this guide helps you understand the hidden layers behind the chatbot."

After: "ChatGPT gets its information from two places: a massive pre-trained dataset with a knowledge cutoff, and real-time web browsing via ChatGPT Search. Understanding which one is active matters a lot for how you interpret its answers, and for whether your content has a chance of being cited."

The before version is a table of contents dressed up as a paragraph. No LLM is citing "this article breaks down." The after version is a self-contained factual claim a model can extract and quote directly.

Example 3: Generic agency intro

Before: "In today's competitive digital landscape, having a strong online presence is more important than ever. Businesses that invest in SEO see better rankings, more traffic, and ultimately more revenue."

After: "Most businesses rank for fewer than 10% of the keywords they should own. The gap isn't authority — it's content architecture. Here's how to audit what you're missing and build a plan to close it."

Nobody is citing "having a strong online presence is more important than ever." It contains no information. The after version opens with a specific, citable claim.

The Intro-Check Prompt

I've been running this in Claude before publishing anything this week:

You are an intro paragraph auditor for AI search optimisation.
Score the following intro out of 10 across five dimensions:

1. BLUF (Bottom Line Up Front): Does sentence 1 contain a self-contained
   factual claim that directly answers the primary query? (2 points)
2. Entity density: Are the key entities (topic, brand, location) present
   within the first 50 words? (2 points)
3. No preamble: Does it avoid openers like "In today's...",
   "Have you ever...", "It's important to..."? (2 points)
4. Human hook: Does it give a reader a reason to continue via a stat,
   bold claim, or clear problem? (2 points)
5. Standalone coherence: If extracted alone, does it make complete
   sense without the rest of the article? (2 points)

For each dimension: score, one sentence of diagnosis, one specific fix.

Intro:
[paste here]

Anything under 7 gets rewritten before it goes live. Running this on three older articles this week surfaced the same pattern every time: good hook, zero BLUF, low citability.

If you want to go deeper on how AI search engines evaluate content structure, the AI search ranking factors breakdown covers what the studies actually show.

This Week's Takeaway

Your intro has two readers: the human deciding whether to stay, and the LLM deciding whether to cite. The hook is handled. The BLUF isn't. One sentence at the top of every article — a self-contained factual claim that answers the query directly — is the highest-leverage structural change you can make for AI search visibility right now.

Frequently Asked Questions

Does BLUF hurt readability for human readers?

No, and the data backs this up. Hostinger's experiment found citation rates improved without any drop in engagement metrics. BLUF intros also convert better with humans: readers who find their answer immediately in sentence one are more likely to keep reading for depth, not less. It mirrors how people read Wikipedia: the lead paragraph states everything, and they scroll for context if they want it.

How long should a BLUF sentence be?

20-50 words is the practical target. Long enough to be substantive and self-contained, short enough to be extractable as a single chunk. Test it: copy sentence one into ChatGPT and ask "does this answer [query]?" If the answer is no, it's not BLUF yet.

Should every page have a BLUF intro?

Every informational page, yes. Product pages and transactional pages follow different intent patterns. For any page targeting an informational query — how to, what is, guide, explained — BLUF is the standard if AI visibility matters to you. Given AI Overviews appear on 99.9% of informational queries, that's most of your content inventory.

Want more? Follow the full AI Search Tracker or explore the AI SEO hub.