PublishedVerifiedEvery 6 weeksSources5 namedAuthored bySquareRank Team
AI Overviews · § 1.3.3 · How-to
First-200-Word Passages for AI Overviews
AI Overviews extracts from passages, not full pages. The pattern that gets cited is consistent: a bolded one or two sentence answer at the top of the page (and at the top of every H2), expanded into context, named sources, and evidence. Search Engine Land's 2026 GEO research1 names the pattern as "clear, direct answer at the top of every section." Secondary sources2 add a working word-count band of roughly 134 to 167 words per self-contained passage — long enough to carry a full thought, short enough to fit a citation card.
This leaf is the Squarespace 7.1 application: block-by-block on a default blog template, on the homepage and service templates, and on the platform's harder cases. The 134-167 word band is a working guideline drawn from secondary 2026 sources, not a published Google rule — treat it as a directional target, not a magic number.
§01The short answer
TL;DR — bolded lead, 134-167 words, repeat under every H2
The first 200 words of a Squarespace blog post should answer the page's primary query directly, in a self-contained passage that holds up if pulled out of the page and read alone. Bold the passage. Aim for 134 to 167 words. Then expand with context, named sources, and the technical depth that distinguishes you from generic competitors. Repeat the same pattern at the top of every H2 section, because AI Overviews extracts at the section level as well as the page level. The 134-167 band is a working guideline from secondary 2026 sources, not a Google-published rule — treat 150 words as the centre and don't lose sleep over hitting the exact endpoints.
The discipline is the change. Most Squarespace blog posts written before 2024 lead with a personal anecdote, a quote, or a scene-setting paragraph — format conventions that worked for human readers and made the article feel warmer. The 2026 reality is that AI Overviews reads the opening 200 words first and decides extraction from there; an anecdote at the top means the answer arrives at paragraph three, which means the answer does not get cited. Bury the anecdote in paragraph four, lead with the answer.
§02The mechanism
What passage extraction actually does
Google's Search index since 2020 has supported passage-level ranking — the engine surfaces specific passages from longer pages when those passages answer the query better than the page as a whole. AI Overviews builds on the same mechanism: Gemini reads candidate pages from the top of the SERP, identifies the passages that most directly answer the query, and synthesises an answer from one or several of them. Pages that lead with the answer offer the easiest extraction target; pages that bury the answer make the engine work harder, and harder-to-extract content gets passed over.
Search Engine Land's 2026 GEO research1 frames the pattern from the editorial side: "AI engines break pages into individual passages" and choose between them based on how directly each passage answers the underlying query. The implication is structural: the passage you want quoted has to look like a citation card before the engine ever looks at it. A bolded lead at the top of a section is recognisable as a candidate passage at a glance; a buried answer somewhere in the body is not.
The pattern is not new editorial advice. Direct-response copywriting, technical documentation, and reference encyclopedias have used "answer first, expand second" for decades. What changed in 2024-2026 is that the same pattern now feeds an automated extraction layer that decides AI citation. The work is the same; the audience now includes the engine.
The 2026 passage landscape, in numbers
~48%
of tracked Google queries trigger an AI Overview as of February 2026 (BrightEdge).
The opening passage of a page is where the engine starts reading, and it carries disproportionate weight in how the page is summarised, classified, and cited. AI Overviews extraction reads the title, the H1, and the first 200 words first; if those layers do not unambiguously identify what the page answers, the engine forms a less confident model of the page and weighs it lower as a citation candidate. The fix is straightforward: write the opening 200 words as if they were the only words anyone would read, then write the rest of the page on top of that foundation.
The "first 200 words" framing is heuristic rather than precise. Different engines weight different windows: some look at the first paragraph, some the first section, some the first 1,000 characters. The shared property across engines is that earlier content gets more attention than later content. Optimising for the first 200 words is the durable bet because it works under every weighting variant tested in the field.
For a Squarespace blog post, the first 200 words usually corresponds to the first text block under the post title, plus the opening of the second block. Author this real estate deliberately. A common revision pass: copy the first 400 words into a separate document, identify the sentence that most directly answers the primary query, and rewrite the opening 200 words to lead with that sentence. The rest of the page reorganises around the new opening.
§04The band
The 134-167 word band, and how rigid to be about it
The 134-167 word band is a working guideline drawn from secondary 2026 GEO research, not a Google-published rule. It exists because AI citation cards tend to surface roughly 130-170 word excerpts, and passages authored in that range are more often quoted in full versus truncated. Treat 150 words as the centre and the endpoints as soft. A 110-word passage works if it carries a full thought; a 200-word passage works if every sentence pulls weight. Obsessive word-counting hurts the prose more than it helps the extraction.
The pragmatic discipline is "make it work as a citation card." Read the passage out of context. Does it answer the question? Does it cite or reference its evidence? Does it stand alone? If yes, the word count is in the right range. If the passage needs a preceding paragraph to make sense, it is not yet citable.
§05On 7.1 blog templates
Applying the pattern on a Squarespace 7.1 blog template
On a default 7.1 blog post template, the structure that ships is: post title (H1), feature image (optional), body region with text and image blocks. The 134-167 word opening passage lives in the first text block under the title, before any feature-image block. Each subsequent H2 — added as a text block tagged H2 in the editor — opens its section with the same bolded-lead pattern. Squarespace's heading tag dropdown lets you toggle H2/H3 inside a text block, so the pattern works in the native editor without code injection.
One Squarespace-specific quirk worth flagging: the platform's auto-generated post excerpt is taken from the first text block by default. If your bolded opening passage starts with a bolded sentence, the auto-generated excerpt also starts with the bolded sentence, which can look odd in the blog index card. The fix is to set a manual excerpt in Post Settings → Options → Excerpt that mirrors the opening sentence without the bold formatting.
A second quirk: Squarespace 7.1 ships a "Read More" excerpt break feature that can be inserted mid-block. Place the break immediately after the bolded opening passage. The blog index card then shows the entire bolded answer as the preview, which serves the same passage discipline at the index level.
MarkdownThe pattern, expressed as the Markdown block you might paste into Squarespace's Markdown block
## Why does my Squarespace blog post not appear in AI Overviews?**The most common reason is that the post answers the query somewhere in the body,but not in the first 200 words. AI Overviews extraction reads the opening passagefirst; pages that bury the answer in paragraph three offer a harder extraction targetand get passed over for pages that lead with it. The fix is editorial: identify thesentence that most directly answers the primary query, move it to the opening, andexpand it into a 134-167 word self-contained passage. Then reorganise the rest ofthe post around the new opening.**The discipline is not new; technical documentation has used "answer first, expandsecond" for decades. What changed in 2024-2026 is that the same pattern now feedsan automated extraction layer that decides AI citation...
Reference the sibling leaf on heading hierarchy if your blog template ships multiple H1 elements per page — the passage discipline assumes one clear H1 per page. Two clean H1 elements make passage extraction harder, no matter how disciplined the opening passage is.
§06On non-blog templates
Applying the pattern on service pages, homepage, and product pages
The 134-167 word opening passage works on every page type, not just blog posts. On a service page, the passage replaces the hero subtitle's role and lives just below the H1, before any feature grid or testimonial section. On the homepage, the passage answers the brand's primary question ('what does SquareRank do?') in self-contained form. On product pages, the passage answers 'what is this product, and who is it for?' before the price and image gallery.
The hardest case is the marketing homepage built around a hero composition with a tagline and a call to action. The tagline is typically too short to function as a 134-167 word passage; the natural fix is a second text block immediately below the hero that holds the citable passage. Visually it can sit small — the citation card does not need the same prominence as the hero — but it needs to exist as readable text on the page, not just structured into metadata.
§07Anti-patterns
What not to do (and why)
Three anti-patterns kill passage extraction even when the rest of the install is clean. First: leading the page with a personal anecdote or scene-setting paragraph that delays the answer. Second: hiding the answer inside an accordion, tab, or 'click to expand' interaction that requires user input to reveal. Third: making the opening passage so generic that it could appear on a hundred other pages and the engine can't decide why your version should be cited. The fixes are editorial, not technical.
The accordion case is the one most Squarespace owners do not realise affects them. Some 7.1 templates default to showing FAQ content inside collapsible blocks, with the answer hidden until clicked. The visible-content requirement in Google's structured-data policy means the answers should be in the DOM (and they usually are, just visually collapsed), but extraction still favours content that is visible without user interaction. The fix on FAQ pages: keep the answers open by default, or use a different block type for the citable content.
The generic-opening case shows up most often on pages built from a template purchased on Etsy or a similar marketplace. The opening passage was written to fit "any business" and has not been customised to fit yours. Two minutes with the editor and a more specific opening — one that names what you do, who you do it for, and what makes your version different — produces a passage that is both more useful to human readers and more citable to AI engines. The work is one investment in two audiences.
