The seven signals that predict citation
Cited pages share seven properties, checked in the order below because each one gates the next: a page that fails signal seven never gets far enough to be judged on signal one.
01 Why do quotable, self-contained sentences get cited?
Short sentences that state a complete claim get lifted into answers because a model cannot safely edit a longer passage down without changing what it says. If your key claim is the fourth clause of a ninety-word paragraph that opens with "this also means," it cannot be extracted cleanly. A passage is a short, self-contained span of text that a model can quote without rewriting it, and passages that survive extraction have their own subject and no pronouns pointing backward.
Compare two sentences. "As we've discussed, this typically takes around a fortnight in most cases" is unusable outside its paragraph. "A GeoTest Site Analysis usually returns a report in one to three minutes" is a sentence a model can drop into an answer with your name after it.
Most sites miss this. In GeoTest's benchmark of 4,999 US business sites, scanned between 11 and 13 October 2026, 96.8% of homepages had sections that do not open with the answer.
Do this: open every important section with one sentence that states the claim in full. One idea per sentence, no "this," "that," or "the above" in the first line.
02 Why do numbers outperform adjectives?
Numbers outperform adjectives because a model comparing several vendors can only describe concretely the one that gives it a figure to repeat. "Fast shipping" is not a fact; it is a mood. "Ships in two business days" is a fact, and it is the kind of statement an answer needs. The same applies to price: a page that says "contact us for a quote" gives the model nothing to work with when a user asked "how much does X cost." A page with a starting price, even a range with conditions, becomes the answer instead.
This is the single biggest category in the rubric. Citation Worthiness counts concrete facts per 1,000 words: 12.5 or more earns the full fact-density points, roughly one checkable fact every 80 words, and facts with their source beside them earn more again.
The same benchmark found that 91.5% of homepages carry too few concrete facts, and 88.6% miss both this check and the answer-first one.
Do this: take the three adjectives you are proudest of and replace each with a number you can defend, then put those numbers in their own sentences, not inside a marketing paragraph.
03 Why does structured data matter for citation?
Structured data matters because it tells a machine what a page is without requiring it to infer the answer from prose. Schema.org markup, written in JSON-LD, states in machine-readable terms what a page is, who published it, and what questions it answers. It is the difference between a parser guessing your business name from a logo and reading it from a field labelled name.
Structured Data is worth 12 of the GeoTest rubric's 100 points in v6.0. It is weighted there not because markup is magic, but because it is the cheapest unambiguous signal available: an afternoon of work that removes an entire class of misreading. Organization, WebSite, Article or BlogPosting, Breadcrumb, Product where relevant, and FAQPage cover most sites. Many sites already have some of it: in GeoTest's scans of 5,278 homepages between 4 and 27 October 2026, 79.0% carried some JSON-LD, 55.2% marked up an Organization or LocalBusiness, and 27.2% a BreadcrumbList.
Do this: add Organization and WebSite sitewide, then the page-type node on each template. Validate with a structured data testing tool before you move on.
04 Why do AI engines prefer a visible, checkable author?
AI engines prefer content signed by a named, checkable author because attribution is a small risk the engine takes on when it puts a name next to a claim. Anonymous content reads like an anonymous tip. A named author with a real bio, an About page that establishes why that person would know, and an Organization node tying the site to a real company all give a model a reason to attribute rather than paraphrase.
Named authorship is one trust signal; factual consistency across independent sources is another. When your own pages, your directory listings and other sites state the same facts about you, an engine has less reason to doubt any one of them, and accuracy on your own page is what it checks those sources against.
This is the part most B2B sites skip and most publishers get right. If your best explainer page is signed "Admin" or nothing at all, you are asking to be quoted without credit at best.
Do this: byline every substantive page, link the byline to an author page, and connect both to your Organization schema.
05 Why does depth on one page beat thinness across ten?
One page that answers a question and its eight follow-ups gets pulled into more answers than eight pages that each answer a single question badly, because every follow-up is a separate query that needs a document to find. A cornerstone page is a single page built to cover one topic and its common follow-up questions completely, so that many different queries can retrieve it.
The trap is padding. Length with nothing in it does not help, and in the v6.0 rubric it scores nothing at all: a model is not rewarded for six hundred words of throat-clearing and neither is the page. What helps is more answerable specifics: prices, steps, constraints, exceptions, definitions, comparisons.
Do this: pick your single most commercially important topic and build one cornerstone page that covers it completely, then link the thin pages into it rather than letting them compete with it.
06 Why does visible freshness act as a tiebreaker?
Visible freshness acts as a tiebreaker because, when two sources support the same claim, a recent and legible date is a reasonable way for a model to choose between them. Models are told a lot about recency by the page itself: a published date, a "last updated" line, current years in the copy, and a dateModified field in schema. In v6.0 that is its own category, worth up to 7 points on articles, docs and product pages: 5 for a recent update and 2 when the visible date and the schema date agree.
Be honest about it. Changing a date without changing the content erodes trust with both readers and engines. Update the substance, then update the date.
Do this: put a real published and updated date on every article, mirror it in schema, and set a calendar reminder to revisit your top five pages quarterly.
07 Why does server-rendered text outrank everything else on this list?
Server-rendered text outranks every other signal because it decides whether the page is even readable before any of the other six are judged. If your site is a client-rendered React, Vue, or page-builder app, the server may return a near-empty HTML shell that only becomes a page once the browser executes the bundle. Crawlers differ in whether they render JavaScript, how long they wait, and how much they retry. Google's own JavaScript SEO documentation has flagged rendering dependencies for years, and Google is the most patient crawler in the set.
The AI crawlers are less patient still. In a December 2024 analysis, Vercel and MERJ found that OpenAI's GPTBot made 569 million requests across Vercel's network in one month and Anthropic's crawler 370 million, and that none of the major AI crawlers render JavaScript. In GeoTest's benchmark, the main text needed JavaScript to appear on 37.1% of homepages and on 88.8% of the 233 Lovable sites.
The safe standard is simple: your important content should be in the server response. Server rendering, static generation, or a prerendered fallback all satisfy it, and browser-based audits cannot catch this failure because they run the JavaScript before they look.
Do this: fetch your own page the way a crawler does, with JavaScript disabled, and read what comes back. If the headline is missing, fix that before anything else on this list.
What does published research say about citation?
Published research points to the same properties as the seven signals, and it keeps finding that specific facts beat length.
- The GEO study. In the GEO study by Aggarwal et al. (Princeton and others, published in 2024 at KDD), adding statistics, source citations and quotations to a page raised its visibility in AI answers, with a 41% lift from quotations, while keyword stuffing did little.
- Facts, not words. Stacc, across 57,253 URLs and 1.85 million fact appearances, found that pages with 12.5 or more facts per 1,000 words were 4.2 times more likely to be cited. Ahrefs, across 560,346 AI Overviews and 174,048 pages they cited, found no advantage for long pages.
- Schema is less common than assumed. Semrush, across 5 million cited URLs, found Organization schema on only 25% to 34% of the pages cited.
- The answer often replaces the click. A Pew Research Center study of 900 people in the US found that users who saw a Google AI summary clicked a traditional result in 8% of visits, against 15% for users who did not.
The Stacc, Ahrefs and Semrush figures are the studies the GeoTest methodology cites for moving points from length to facts in 2026.
How do the seven signals map to the GeoTest rubric?
The seven signals distribute across the six weighted categories in the v6.0 GeoTest rubric, and no single category earns points on its own if the crawler cannot extract the page first.
| Category | Points | Signals it captures |
|---|
| Citation Worthiness | 34 | Fact density, coverage of agreed facts, attributed quotes, external sources, quotable definitions |
| Answer Structure | 22 | Answer-first sections, heading hierarchy, a visible Q&A block, FAQPage schema backed by visible text |
| Authority and Trust | 17 | Social proof, named authors, contact methods, privacy and terms links |
| Structured Data | 12 | Schema that fits the page, parses and agrees with the visible text |
| Extractability | 8 | No noindex, text in the raw server response, canonical tag, viewport |
| Freshness | 7 | A visible date that matches dateModified in schema |
Extractability carries the fewest points and matters the most. A site that fails it usually fails four other categories at the same time, because the crawler never saw the content that would have earned those points.
What doesn't GEO scoring claim to know?
GEO scoring does not claim to see inside any engine's ranking process, only to model the mechanism that produces citations. What is observable is which sources appear in answers, what those pages have in common, and what changes on a page when you change it. The seven signals above are a model of that mechanism, not a leak from one.
The weights in the GeoTest rubric are a considered judgment about which properties matter, published openly on the methodology page so anyone can disagree with them specifically. They are not derived from a survey, and this article does not present them as one.
What is the one-week plan to fix these signals?
Fixing all seven signals takes about a week of focused work, followed by a re-test thirty days later once the crawler has revisited the page.
- Day 1, check what the crawler receives. Run a scan, or fetch your top pages with JavaScript off. Fix the raw response first if it is empty.
- Day 2, schema. Add Organization and WebSite sitewide, plus the correct page-type node on your templates, then validate.
- Day 3, bylines and About. Name the author, write the bio, link them together.
- Day 4, rewrite openers. The first sentence of every section becomes a standalone claim.
- Day 5, put the numbers in. Price, timeline, coverage, capacity, whatever your buyers ask.
- Day 6, add the questions. Eight real buyer questions with direct answers and FAQPage markup.
- Day 30, re-test. Structural fixes should be visible by then. If nothing moved, check what the crawler actually received.
See which of the seven signals your site is missing.
GeoTest's own homepage carries a live badge showing its current score on this rubric, scored the same way this article recommends scoring yours. GeoTest fetches your pages the way an AI crawler does and scores all six v6.0 categories against the published rubric.
Run your free GEO test →Frequently asked questions
Short answers to common questions about how AI engines decide what to cite.
How do AI engines choose which sources to cite?
AI engines retrieve a set of candidate pages, extract passages that answer the question, and attribute the ones they use. Citation is won at passage level, not page level. A page gets cited when it contains a short, self-contained, factual statement that answers the question directly, on a page the crawler could actually read, from a source it has some reason to trust.
Is getting cited by ChatGPT the same as ranking on Google?
No, getting cited by ChatGPT is not the same as ranking on Google. Ranking is a list of ten positions competing for a click. Citation is a handful of sources folded into one answer, usually with no scrolling and no second page. A site can rank well and never be cited, most often because its key facts are buried in prose or absent from the server response.
Does word count affect citation?
Depth helps, but word count alone does not, and in the v6.0 GeoTest rubric length scores nothing at all. Two gates remain for articles, guides and other content pages: under 150 words of body text caps the page at 40, and 150 to 299 words caps it at 70, because there is not enough there to quote. Above that, a dense 600-word page beats a padded 2,000-word one.
Do you need schema markup to be cited?
No, you do not need schema markup to be cited, but it removes guesswork. Schema states in machine-readable terms what the page is, who published it, and what the questions and answers on it are. In the v6.0 GeoTest rubric, Structured Data is 12 of 100 points, because it is the cheapest way to make a page unambiguous.
How many points does structured data carry in the GeoTest rubric, and why that much?
Structured Data carries 12 of the rubric's 100 points in v6.0. It sits there because it is the cheapest unambiguous signal available: an afternoon of work with Organization, WebSite, Article, Breadcrumb and FAQPage markup removes an entire class of misreading that would otherwise cost points across other categories.
What should be the first step in a GEO fix, and why start there?
The first step is checking what the crawler actually receives, by fetching your top pages with JavaScript disabled. It comes first because it gates everything else: if the raw server response is empty, no amount of work on quotable sentences, schema or authors will help, since the content was never in the candidate set.
How long does it take for changes to show up in AI answers?
How long changes take depends on when each engine next fetches your pages, which varies by engine and by site. Structural fixes such as server-rendered text and schema are detectable on the next crawl; authority signals from other sites accumulate over months. Re-test roughly every 30 days rather than daily.
What is the single most common reason a site is never cited?
The single most common reason is that the crawler did not receive the content. Client-rendered pages can return an HTML shell that only becomes a real page after JavaScript runs, and crawlers differ in whether and how long they render. If the raw server response is empty, nothing else on this list matters.
