PublishedVerifiedEvery 6 weeksSources5 namedAuthored bySquareRank Team
ChatGPT · § 1.1.3 · How-to
Write a Squarespace Blog Post ChatGPT Will Actually Cite
Every H2 on a citable page opens with a bolded one to two sentence lead of 134-167 words2 that answers the section's question without requiring a click. The rest is expansion: named sources, dates, statistics, technical depth. The shape works because that is the band AI engines extract from most reliably, and the format ships natively on Squarespace 7.1 blog templates without code injection.
This leaf is the block-by-block writing pattern for a Squarespace 7.1 blog post that ChatGPT will quote. The rule is short — open every section with a self-contained answer — but the Squarespace-specific application has details most generic GEO guides skip: how the section-based 7.1 layout fragments your H1/H2 hierarchy, where the Code Injection patch goes, and the text block width that actually lets the passage breathe.
§01The rule
The 134-167 word passage rule
A citable passage on a 2026 AI-search page sits between 134 and 167 words. Long enough to carry a full thought — definition, qualification, one piece of evidence. Short enough to fit a citation card on a ChatGPT answer screen or under an AI Overview block. The rule is not a hard limit; ChatGPT will quote a 110-word passage if it is the cleanest answer available, and a 200-word passage if nothing tighter exists. But the centre of the band is where extraction consistently lands.
The pattern shows up in the practitioner research2 across multiple 2026 GEO guides: AI systems favour self-contained passages in that band when deciding which span to lift into an answer. Search Engine Land's broader GEO guide1 frames the same shape in different language — "start each section with a clear, direct answer, then expand with context" — and the length recommendation is the practical implementation of that frame.
The Squarespace-specific application is what makes this leaf necessary. Most generic GEO advice assumes a WordPress or Webflow blog where the author has full control over heading semantics. Squarespace 7.1's section-based layout introduces edge cases — fragmented H1 hierarchy, text blocks that span the wrong width, code-injection placements that compete with the auto-emitted Article schema — that change how the passage rule needs to be applied.
The 134-167 band, calibrated
134
lower bound of the extraction-favoured passage band in 2026 GEO research.
A citable Squarespace blog post follows a tight pattern. H1 with the primary query restated naturally. 50-100 word intro paragraph that previews the answer. Five to seven H2 sections, each opening with a 134-167 word bolded lead, each expanded with 200-400 words of context and named sources. One H2 reserved for FAQ or 'frequently asked' format. One callout per major section. One named statistic per section minimum. Author byline pointing at a Person schema entity. Total length 1,500-2,500 words for editorial topics, 800-1,500 for technical how-to.
The intro paragraph carries a separate function: it is what surfaces in Google's snippet, in social shares, and sometimes as the opening line of a ChatGPT answer when the model decides the whole-page summary is the right format. Keep it neutral and definitional — what is this page about, what question does it answer, what is the takeaway. Save the sales tone for further down or for a sidebar wedge; first-200-words shaping for citation extraction beats first-200-words shaping for persuasion.
The named-sources-per-section minimum is what separates a citable page from a generic one. A page making five claims with five named publishers, dates, and primary-source links reads, to the model, like a brief technical document. A page making the same claims without attribution reads like a marketing post. The model picks the technical document. Squarespace's own help center3 says the same thing in different language: "support with statistics," "original content with unique insights," "clear headings and subheadings."
§03The blocks
Block-by-block on Squarespace 7.1 blog templates
On a Squarespace 7.1 blog template, the post body lives inside a single content area that accepts text blocks, image blocks, code blocks, and Markdown blocks. The passage rule maps cleanly: one text block per H2 lead (bolded, 134-167 words); one text block per expansion paragraph; one Markdown block for any code or schema snippet; one image block per visual aid. The default text block width is correct for citation extraction — somewhere between 60 and 75 characters per line. Wider blocks (full-bleed text) can fragment what the AI parser sees as a single passage.
The single-content-area constraint matters because it limits how much template-level decoration sits between your passages. Squarespace's auto-emitted blog post HTML wraps each block in semantic article markup, which is helpful for AI extraction — the parser knows where each passage starts and ends. The interference comes from full-bleed image or quote blocks injected mid-passage; they break the parser's read of a clean lead-plus-expansion span.
One practical detail worth knowing: when you paste from a Google Doc or Word into a Squarespace text block, the rich-text editor preserves font-family overrides and inline styles that increase HTML complexity. The result is parseable but noisier than ideal. Pasting into a Markdown block, or into the regular text block with "Paste as Plain Text" (Shift+Cmd+V on Mac, Shift+Ctrl+V on Windows), strips the noise and produces cleaner output.
§04The hierarchy
Heading hierarchy and the 7.1 section break
Squarespace 7.1 introduces 'sections' as a layout primitive, and each section can declare its own heading. On a blog post template the sections are usually flat, but on a regular page they often produce fragmented hierarchy — H1 in the page header, another H1 inside a section, H2 buried under an H3 in another. AI parsers read fragmented hierarchy as a signal that the page is not well structured, which lowers extraction confidence. The fix on a blog post is to keep all headings inside the post body and use H2/H3 only; the fix on a regular page is the Code Injection patch in the next section.
The blog post case is simpler. Open the post in the editor, add headings via the format dropdown in a text block, choose Heading 2 for top-level sections and Heading 3 for sub-points. Avoid Heading 1 inside the body — that role is owned by the post title in the template header. Avoid Heading 4 entirely unless you genuinely need a fourth level; AI parsers rarely read past H3, and a clean two-level hierarchy outperforms a four-level one for extraction purposes.
The regular page case is where Squarespace 7.1's section-based layout creates problems. Each section in the page builder gets its own heading slot, and the heading level inside a section is set per-block — which means a 7.1 page can end up with three H1s (page header plus two section headers), with H2s buried unevenly inside. The Code Injection patch in the next section forces the hierarchy back to one H1 and well-ordered H2/H3 beneath it.
§05The sources
Named sources, dates, and statistics inline
Replace every vague attribution with a named publisher, a date, and a primary-source link where possible. Two named sources per page minimum on editorial content; three or more on technical content. Sources should be 2026-current — anything older than 18 months should be flagged or refreshed. The format AI engines extract from looks like: 'According to Search Engine Land's 2026 GEO guide, ChatGPT serves 800 million weekly users.' Compare with the rejected version: 'Studies show ChatGPT is huge.'
The mechanism is hallucination-prevention on the engine side. When a model decides whether to cite a passage, it weighs the cost of repeating an unverifiable claim against the value of the answer. Named attribution lowers that cost — the model can defend the citation if a user asks where the number came from. The engine's training has explicitly biased it toward attribution-heavy prose because that prose is safer to surface.
On Squarespace specifically, the constraint is that you cannot install a Footnote plugin or a Sources sidebar widget without code. The pattern that works without code: a manifest box at the foot of the article (a single Markdown block with numbered links), inline citations using a simple (Publisher, 2026) parenthetical style, or — if you have Business plan or above — the right-rail SourceManifest pattern this site uses, installed via Code Injection.
§06The patch
The Code Injection patch for H1/H2 hierarchy
For regular pages where 7.1's section-based layout fragments your heading hierarchy, a small Code Injection snippet in the page header forces the right structure. The pattern: hide any duplicate H1s in section headers via CSS, then add Article JSON-LD schema in the page header that declares the canonical headline and sub-headings. Together they tell parsers (both AI and traditional) which heading is the page-level H1 and which are subsections.
JSON-LDArticle schema with canonical heading hierarchy — paste into Page Settings > Code Injection > Header
<script type="application/ld+json">{"@context":"https://schema.org","@type":"Article","headline":"The H1 you want AI engines to read","datePublished":"2026-05-18","dateModified":"2026-05-18","author": {"@type":"Person","name":"Your Name","url":"https://yoursite.com/founder/"}}</script><style>section[data-section-id]h1:not(.entry-title) {display:none;}</style>
The Article schema5 tells AI parsers which headline is canonical even if the rendered HTML contains duplicates. The CSS rule hides the duplicate H1s visually and removes them from the accessibility tree, which helps both AI extraction and screen-reader behaviour. Code Injection is locked to Business plan and above on Squarespace; on Personal plan the workaround is to avoid full-page templates with multiple sections and use a single-section page or a blog-post layout instead.
§07The per-post check
The seven-item per-post checklist
Before publishing any citation-target page, run the seven-item check below. The shape compounds: each item is small, the cumulative effect is large. A page with all seven boxes ticked is materially more likely to be cited than a page missing two or three. The full 22-item version including crawler, schema, and measurement items lives in the dedicated checklist leaf.
Lead length. Every H2 opens with a bolded 134-167 word self-contained answer.
Source density. 2+ named publishers with dates inline, minimum.
Statistics. 1+ specific numerical claim per major section, with the source named adjacent.
Heading hierarchy. One H1, H2 for sections, H3 for sub-points. No H4+ unless essential.
Author wiring. Article JSON-LD schema points author.url at the Person entity on /founder/.
Dates.datePublished and dateModified are current ISO-format strings.
Internal links. 2-5 contextual links per 1,000 words — to the cluster hub, to the pillar, to sibling pages.
