FREE CLAUDE SKILL · AEO CONTENT WRITING
Write blog posts that AI actually cites.
Answer engines do not read your post top to bottom. They split one question into a dozen sub-questions, pull the passages that answer each one, and build a reply from the pieces. This free Claude skill writes for that. It maps the fan-out, structures every section to survive being lifted out of context, and checks the draft before you publish. Built by Lil Big Things, the GEO agency for B2B tech.
Runs in Claude · Connects to Ahrefs and Exa if you have them, works fine without · No card required
Why this exists
Ranking first stopped guaranteeing you get quoted.
Your post can sit at position one and still never appear in the answer a buyer reads. The share of top-10 pages cited in AI Overviews fell from roughly 76% in 2025 to roughly 38% in 2026 (ALM Corp, directional). Rank and citation have come apart.
The reason is mechanical. Engines rarely search your exact query. They fan it into 8 to 16 sub-queries for anything complex, run them in parallel, retrieve passages rather than pages, then synthesise one answer from the best pieces. No single page wins the whole query. The page that covers the most sub-queries wins the most of it.
Most content is written for one keyword. That matches a fraction of the retrieval surface.
sub-queries fired for a single complex prompt.
ChatGPT runs 4 to 20. Deep-research modes go far higher. Ahrefs watched one task fire 420 searches. Pages that rank for the fan-out sub-queries see roughly a 161% lift in citation likelihood (Surfer SEO, directional).
The seven fan-out types
One question becomes twelve. The skill writes for all of them.
Search patents describe how a query gets decomposed into sub-query types. The skill uses those seven types to build the sub-query map before a word gets written. Each one becomes an H2 with its own self-contained answer.
Related topics
Adjacent context the reader needs next. For a post on the best CRM software, the engine also pulls CRM implementation and data migration. Cover these and your post matches more retrieval requests.
Implicit questions
The concerns nobody types but everybody has. Pricing for small teams. How long setup takes. These sub-queries never appear in keyword tools but they appear in answers regularly.
Comparative
Side by side. Salesforce vs HubSpot vs Zoho. This is the format AI engines cite most heavily. A post that handles the comparison in its own section beats one that avoids it.
Recency
Time-sensitive versions. Engines stamp the current year onto most sub-queries automatically, so an unstamped page misses the query that actually ran. The skill adds year stamps as part of the draft.
Reformulations
Same intent, different words. Customer management tools. Sales pipeline software. A post that only optimises for one phrasing misses every reformulation.
Contextual
Segment, industry, location. CRM for startups. CRM for healthcare. These narrow sub-queries often have low competition and high citation rates because most pages ignore them.
Next-step
What the reader asks immediately after. How to import contacts into a CRM. A post that anticipates the next question keeps someone in your content and extends how much retrieval surface you own.
Cover the fan-out and you match more of the retrieval surface. Cover one keyword and you match a sliver of it.
What makes a passage citable
Engines lift passages, not pages.
Retrieval happens at passage level. A model pulls a self-contained chunk out of your post and drops it into an answer. If that chunk only makes sense with the paragraph above it, it gets discarded. If the engine cannot work out which company a claim refers to, it gets discarded.
Anatomy of a citable passage
ENTITY WELDING · why passages get dropped
Dropped: "Its filet mignon is the best in the city."
Engine cannot resolve "its" or "the city".
Cited: "Boucherie Union Square's filet mignon is the
best in New York City."
Entity welded to the claim, same sentence.
ANSWER CAPSULE · the shape of a citable section
H2: A question phrased the way a person asks it
Line 1: The direct answer, 40 to 60 words
Then: Expansion with named tools, real numbers, dates
Test: Cut it out. Does it still answer the question?Top-cited content format by engine (Profound, 2026, directional)
| Content type | Claude | ChatGPT |
|---|---|---|
| Listicles | 36.4% | 19.7% |
| Brand / product pages | 17.6% | 21.2% |
| Blog / opinion | 13.2% | 7.2% |
| Forum / UGC | 0.9% | 15.8% |
| Wikipedia | 0.6% | 2.5% |
If your buyers research in Claude, ranked lists and clear definitions do the heavy lifting. If they are in ChatGPT, community presence matters far more than most B2B teams assume. The skill asks which engines matter before it picks a format.
What makes it different
Built by an agency that publishes this way.
Researches before it writes
Pulls keyword volume and difficulty from Ahrefs, and real sub-queries from Exa, so the fan-out map reflects what people actually ask rather than what sounds plausible.
Checks the cluster first
Looks for pages you already have on the topic. Two pages targeting one sub-query compete with each other in retrieval as well as in rank. The skill flags the conflict before drafting.
Eleven rules, then an eleven-point check
Front-loaded answers, welded entities, question-format H2s, year stamping. The draft gets checked against every rule before it reaches you.
Never invents a statistic
Gaps come back marked, not filled. A fabricated figure in a piece arguing for trustworthiness fails the moment someone checks it.
Hands over distribution
Roughly three-quarters of what AI cites about a brand comes from sources the brand does not own (Omniscient Digital, 23,387 citations, Jan 2026). Publishing is the small part. The skill says where else it needs to go.
Honest about the numbers
Every figure is attributed and flagged as directional. AEO is early and most published data is vendor-reported. The skill says so instead of pretending otherwise.
How it works
From topic to publishable draft in three steps.
Install the skill
Download the file and save it into Claude. It installs once, then triggers whenever you ask for a post, an article, a guide, or a definition page.
Give it a topic
Say what you want to write about and who it is for. It runs keyword and fan-out research, checks your existing cluster, and comes back with a sub-query map before drafting.
Get the draft and the handoff
A markdown file with full frontmatter, plus internal links to add, where to distribute it, and what to measure per engine.
Who it's for
For the people who have to make content earn its keep.
This is for you if...
You publish B2B content and your buyers research in AI chat windows.
You have a blog that ranks and no idea whether anything on it gets cited.
You want a draft that ships, not an outline you have to finish yourself.
You would rather write one thorough page than five thin ones.
This isn't for you if...
You want volume. This writes fewer, deeper pages on purpose.
You expect citations next week. It compounds, and a single post rarely moves a category on its own.
You are not willing to name real sources and real numbers. That is most of what makes a passage citable.
Want Lil Big Things to run the programme and write the cluster? Book a GEO strategy call
What's in the file
Everything is in the box, and it is yours to edit.
The download is one installable skill file plus the readable source, so you can see exactly how each rule is applied and tune the thresholds to your own standards.
aeo-blog-post.skill The installable skill. Save once, run on demand.
aeo-blog-post/
SKILL.md Workflow, the eleven writing rules, self-check.
references/
aeo-principles.md Fan-out mechanics, passage retrieval, entity
resolution, format and freshness signals.
assets/
post-template.md Fill-in-the-blanks starting structure.Honest note. AEO is early and moving fast. Most of the numbers in this skill are vendor-reported and flagged as directional, because that is what they are. The mechanics of fan-out and passage retrieval are well documented. The exact figures will move. Treat the rules as a strong default and check your own citation data rather than trusting anyone's benchmark, including ours.
More free skills
Five skills. Same job, different stage.
SEO + AEO Audit
Checks whether engines can crawl, render, and cite your site.
Homepage Build + Audit
Grades your homepage against six buyer questions and rewrites what is broken.
AEO Blog Post (this skill)
Writes content engineered to get cited.
Content Brief
Builds a brief a writer can execute without a follow-up call.
Meta Description Optimizer
Rewrites title tags and descriptions against the live top 10.
GET THE SKILL
Get the AEO writing skill free.
Drop your email and we will send the download plus a short guide to running your first fan-out map. We will also send the occasional GEO Times update. Unsubscribe anytime.
FAQ