The 134-167 word extractable passage rule

The 134-167 word extractable passage rule says the first paragraph of every article — the prose before any H2 heading — should run between 134 and 167 words and contain a self-contained answer to the article's central question. The rule emerged from cross-audit observation: below 100 words, opening passages lack context for AI assistant synthesis and get extracted less frequently. Above 200 words, AI assistants tend to summarize the passage rather than quote it verbatim, breaking brand attribution. The 134-167 word range is the empirical sweet spot — long enough for substantive context, short enough to encourage verbatim extraction. The rule scores 5 points in the 40-Point AEO Content Geometry Standard, one of the highest-weight dimensions because opening passages get extracted at materially higher rates than any other content in an article. This guide unpacks why the range matters mechanically, the four-sentence structural pattern that scores best, the common authoring mistakes that produce out-of-range openings, and the practical procedure for engineering passages that fit the rule on first draft.

Why does the 134-167 word range exist?

The range emerged from observation, not arbitrary rule-making. Cross-audit testing across 200+ articles produced consistent patterns:

Below 100 words: Opening passages lack enough context for AI assistants to synthesize confidently. The extraction happens, but the model often pads the quoted text with surrounding context or summarizes rather than quotes — diluting attribution.

100-133 words: Better than below 100 but still on the low end. Extraction rates lift visibly, but the passage often gets paired with content from the second paragraph in synthesized outputs.

134-167 words: The empirical sweet spot. Substantive enough to stand alone as a citable answer. Short enough to encourage verbatim extraction. AI assistants quote passages in this range with the highest fidelity and attribution preservation.

168-200 words: Still works but starts hitting summarization behavior. AI assistants increasingly paraphrase rather than quote at this length.

Above 200 words: AI assistants summarize rather than quote. The original brand context dilutes; attribution becomes weaker.

The 134-167 range reflects how the major AI assistants — ChatGPT, Perplexity, Claude, Gemini — balance extractability against summarization heuristics during synthesis. The range may shift as models evolve, but the underlying mechanic (short enough to quote, long enough to be self-contained) is durable.

What's the four-sentence structural pattern that scores best?

Opening passages scoring 5/5 on the 40-Point Standard typically follow a four-sentence pattern. Each sentence does specific work:

Sentence 1: Definitional answer. Lead with the answer to the article's central question. No throat-clearing, no scene-setting, no narrative hook. The first sentence should be the answer a reader would expect from the title.

Sentence 2-3: Specific elaboration with the most important named entities. Unpack the definitional answer with concrete specifics. Include the brand being optimized for, the platforms covered, the central metric or framework. This is where named entity density gets established for the article.

Sentence 4-5: The supporting statistic with date and source URL. Cite the central supporting statistic. Format: [study or source name](https://source-url) — date is part of the citation; URL is hyperlinked.

Sentence 6-7: Signal what the article covers. Brief preview of the article's structure so the reader knows what to expect. Two or three clauses signaling the major sections.

Sentence 8 (optional): Mention the reader's expected outcome. If the article promises specific value (a playbook, a checklist, a framework), this sentence makes that promise explicit.

Total: 6-8 sentences, depending on sentence length. Target 134-167 words total.

The opening passage of What is AEO? on this site exemplifies the pattern at exactly 167 words. The opening passage of The 10-Point AI Citation Framework follows the same pattern. Both scored 5/5 on Phase 4 audit.

How do you measure word count accurately?

Word counting matters because the rule has explicit numeric thresholds. Manual counting is error-prone for prose with hyphenated words, em-dashes, and parenthetical asides.

The reliable counting procedures:

Procedure 1: Copy passage into a word processor. Microsoft Word, Google Docs, or any editor with built-in word count handles edge cases consistently. Select the passage. Read the word count.

Procedure 2: Use a CLI utility. On macOS or Linux: pbpaste | wc -w (copy passage to clipboard, then run). On Windows Git Bash: similar. Handles edge cases predictably.

Procedure 3: Use a node script for MDX files. A simple script that reads the MDX file, extracts the first paragraph before the first H2, and counts words handles automated audits cleanly.

What counts as a word:

• Hyphenated terms ("AI-powered") count as one word.
• Em-dashes don't separate words ("ChatGPT — and Perplexity" is 3 words: ChatGPT, and, Perplexity).
• Parenthetical asides count their contents normally.
• Numbers count as words ("0.334" is one word).
• URLs count as one word each.
• Citation references with hyperlinks ("the SERanking study") count the visible text only — "the SERanking study" is 3 words.

Standard editorial word counters handle these consistently. The numeric threshold (134-167) is tight enough that 2-3 word miscounting matters; pick a counter and use it consistently.

What if the topic doesn't fit a 134-167 word definitional answer?

Some topics naturally produce shorter openings (concise definitional content) or longer openings (multi-part definitions requiring more setup). The rule's flexibility:

Topics naturally producing short openings (sub-134 words):

Pad with additional supporting context. Add a second supporting statistic. Add elaboration on why the topic matters. The goal is not arbitrary padding but adding substantive context that strengthens the opening's standalone value.

If padding feels artificial, the article might benefit from a broader scope — covering the topic plus its immediate context — which naturally requires more opening content.

Topics naturally producing long openings (above 167 words):

Tighten by removing context that belongs in body sections rather than the opening. The opening should answer the central question; deeper context (mechanism, history, implications) belongs in body H2 sections.

Common tightening patterns:

• Remove the second supporting statistic; one is sufficient for the opening.
• Compress multi-clause sentences. "The framework, which we developed across 12 months of cross-client audits and which has been refined across multiple versions, scores ten dimensions" becomes "The framework scores ten dimensions across signals we developed across 12 months of cross-client audits."
• Move the "what this article covers" preview to a single sentence rather than multiple.

If aggressive tightening still leaves the opening above 200 words, the topic may be too broad for a single article. Consider splitting into multiple articles with narrower openings.

What are the common authoring mistakes?

Four patterns produce opening passages that fail the 134-167 word rule:

Mistake 1: Throat-clearing before the answer. "Recently, much attention has been paid to..." or "In today's competitive landscape..." Both add words without adding substance. Cut the throat-clearing; lead with the answer.

Mistake 2: Narrative hook framing. "When Sarah Chen logged into ChatGPT for the first time..." Story openings feel engaging in print journalism but score poorly on AEO. AI assistants extract passages that answer the article's central question; narrative openings answer a different question (what happened to Sarah Chen).

Mistake 3: Defining terms that don't need defining. "Answer Engine Optimization, often abbreviated as AEO, is the discipline of..." The acronym definition wastes words readers searching for AEO already know. Use the acronym after the definitional sentence: "Answer Engine Optimization (AEO) is the discipline of..."

Mistake 4: Including the "why this matters" before the "what it is." Some authors lead with rhetoric ("AI Visibility matters more than ever before") followed by definition ("AEO is the practice of..."). Reverse the order: define first, justify second. The definitional sentence is the citation target.

How do you engineer a passing opening on first draft?

The systematic procedure:

Step 1: Write the article title. The title forms the central question the opening answers.

Step 2: Draft Sentence 1 — the definitional answer. Write the single sentence that best answers the title. Iterate 2-3 times until it reads cleanly as a standalone answer.

Step 3: Draft sentences 2-3 — specific elaboration. Unpack the definition with concrete specifics. Name the most important entities.

Step 4: Draft sentence 4-5 — supporting statistic. Identify the central statistic supporting the article's argument. Cite it with date and inline hyperlinked source URL.

Step 5: Draft sentences 6-8 — what the article covers + reader value. Preview the article's structure. Optionally mention the reader's expected outcome.

Step 6: Count words. Use a reliable word counter. Document the count.

Step 7: Tune to 134-167 range. If below 134: add elaboration or supporting statistic. If above 167: tighten throat-clearing, remove redundant context, compress sentences.

Step 8: Verify the structural pattern. Sentence 1 is definitional. Sentence 4-5 contains sourced statistic with date+URL. Sentence 6-8 previews structure.

Articles written this way typically score 5/5 on the extractable opening dimension on first draft. The audit-refactor loop catches the 10-20% of cases where the opening drifts out of range during writing.

Frequently asked questions

Does the rule apply to landing pages and sales content?

Less strictly. The standard is calibrated for editorial content. Landing pages and sales content have different optimization goals (conversion vs citation) and can use shorter or longer openings depending on conversion-rate testing.

What about articles with multiple paragraphs before the first H2?

The opening passage is the first paragraph specifically. If the article has multiple paragraphs before the first H2, only the first paragraph scores against the rule. Authors should structure articles to put substantive opening content in paragraph 1, with subsequent paragraphs starting after the first H2 break.

Does the rule apply to comparison content?

Yes, with adjusted opening structure. Comparison articles ("X vs Y") should open with a definitional answer about the relationship between X and Y, followed by elaboration explaining when each option fits best. Sentence 4-5 typically cites the central comparison metric. The 134-167 word range applies normally.

What's the rule for FAQ pages?

FAQ pages have a different structure — typically the first content is a Q&A pair rather than prose. The standard treats FAQ pages as a separate case where the opening rule doesn't apply directly. Score FAQ pages on the other 9 dimensions (question H2s become the FAQ Q phrasing, etc.).

Does the rule's specificity vary across AI assistants?

Empirically, the 134-167 range produces strong extraction rates on ChatGPT, Perplexity, Claude, and Gemini. Grok behaves similarly but with more variance (because Grok's retrieval is X-data-heavy and X content has different structural patterns). The rule is calibrated for general AEO/GEO; per-platform optimization may produce minor variations.

---

Companion guides: The 40-Point AEO Content Geometry Standard · How to write content that ChatGPT will cite · Question-format H2s · Named entity density · Source links and date stamps.