Squarespace AI Search Optimization

Three categories of search now bring traffic to small-business websites. Classical search still drives most of it: Google and Bing return a ranked list of links, the user picks one, the site gets a visit. Answer engines — Google's AI Overviews block⁶ and Bing Copilot⁸ — sit above the classical results and synthesise a direct answer, citing three to six sources inline. Generative engines — ChatGPT, Claude, Perplexity, Gemini — answer entire questions in a conversation and cite the sources they drew from, sometimes prominently, sometimes as a footnote.

The same site shows up differently in each. A site that ranks #4 on Google might be quoted at the top of an AI Overview, ignored entirely by ChatGPT, and cited by Perplexity in second position. Each engine reads the page through a different lens, and the engineering decisions Squarespace made about how its 7.1 templates render HTML mean the lens for some engines does not focus.

The shift is not theoretical. Customers searching for a wedding photographer in Asheville, a therapist who takes evening appointments, or a tax attorney who handles estate planning are asking ChatGPT first and Google second — and what ChatGPT recommends is a function of which sites are crawlable, parseable, and citable in the form generative models prefer¹.

Every generic AI-SEO guide tells you to allow GPTBot³, add schema, and write more content. None tells you what Squarespace specifically does behind the scenes — and the gap matters. Four platform behaviours collide with the AI-search playbook in ways that need Squarespace-specific fixes, not generic ones.

Three more platform-specific gaps land in their own clusters: 7.1's section-based HTML break (which fragments H1/H2 hierarchy in ways AI Overviews extraction struggles with), the AI Visibility panel paywall (Advanced plan only, ~$72/mo, monitors one engine of five), and Squarespace's auto-emitted JSON-LD that occasionally conflicts with hand-injected schema.

The framework is sequential — each step depends on the previous one. Unblocking crawlers without installing schema produces noise; installing schema without entity wiring produces anonymous citations; entity wiring without passage-level structure produces pages AI engines can identify but won't quote. Skip a step and the next one underperforms.

The distinction between retrieval bots and training bots is the part most generic guides skip — and it’s the distinction Squarespace’s toggle box gets wrong. Retrieval bots (ChatGPT-User⁴, OAI-SearchBot³, Perplexity-User, Claude-User⁷, Claude-SearchBot) fetch a page on demand when a user asks a live question — these are the bots that produce live citations. Training bots (GPTBot, ClaudeBot, Google-Extended⁶) scrape content to train future model versions — blocking them does not affect current-engine citations. Squarespace’s 26-bot exclusion list is exclusively training-class; the retrieval bots that decide whether you get cited tonight aren’t on it at all.

The AI Crawlers cluster has the full 26-bot exclusion list, the toggle-state audit, the retrieval-bot map, and a free crawler-check tool that audits your live site in 60 seconds.

Schema is not optional in 2026. AI engines use structured data as the fastest path to entity recognition: which site is "Asheville Family Photography", what is the entity behind that site, and is the author credible? A well-shipped JSON-LD graph¹⁰ collapses days of crawl-and-infer work into a single read.

 { "@context": "https://schema.org", "@type": "Article", "headline": "How to be cited by ChatGPT from a Squarespace site", "datePublished": "2026-05-18", "dateModified": "2026-05-18", "author": { "@type": "Person", "name": "Your Name", "url": "https://yoursite.com/founder/" }, "publisher": { "@type": "Organization", "name": "Your Brand", "url": "https://yoursite.com/" } } 

Pillar 3 (Schema & Structured Data for Squarespace) ships the full copy-paste JSON-LD library and the Code Injection placement pattern for each page type.

llms.txt⁹ is the emerging standard for telling AI engines which pages on your site are canonical, what topics they cover, and which ones to prioritise. Adoption is still uneven — not every engine reads it yet — but the ones that do (Perplexity, several enterprise AI-search products) treat it as a primary index.

The Squarespace workaround is documented in detail in the llms.txt cluster. The free llms.txt generator outputs a copy-paste manifest tailored to your business type.

AI engines disambiguate entities by following the sameAs graph. A founder Person entity with five strong sameAs links (LinkedIn, Twitter, a podcast appearance, a published book on Google Books, a Wikidata Q) is far easier for an AI to confidently attribute than an anonymous byline. The result: when a user asks the AI for "the wedding photographer in Asheville who writes about pricing", a recognised entity is more likely to be the cited answer than an unrecognised one.

The 134-167 word passage rule is the most underrated change in AI-search optimisation in 2026. Search Engine Land's GEO research¹ consistently finds that AI engines lift answers from passages of roughly that length — long enough to carry a full thought, short enough to fit a citation card. Pages that lead each H2 with a bolded answer get cited; pages that bury the answer in the third paragraph get read but not quoted.

This pillar uses the format throughout. Notice that every H2 in this article opens with a bolded lead that stands alone — that's not a stylistic choice, it's the citation-hygiene rule applied to the playbook describing it.

The pattern is easy to test. Search ChatGPT for "best practices for [your industry]" and inspect the citations: every cited passage tends to follow a tight format — a clear thesis sentence at the start, named sources inline, dates next to claims, and an answer that holds up on its own. Pages that do not look that way may rank in Google but they will not be quoted in a generative response.

The dark-traffic problem is real and not solvable inside Squarespace alone. The practical workaround is a tracking spreadsheet: a list of 10-15 queries you want to be cited for, run weekly across each AI engine, with screenshots logged when you appear. Pair that with Google Search Console for indexed-page health, and you have a workable measurement loop.

The free SquareRank audit runs this measurement pass for you in 60 seconds and returns three scores: SEO, AEO, and GEO. It's the same scan we run on every $299 install before and after.