> ## Documentation Index
> Fetch the complete documentation index at: https://docs.daftplug.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Answer engine optimization

> What makes a page quotable by ChatGPT, Perplexity, and AI Overviews.

Ranking and being cited are different problems.

A search engine returns a list and lets the reader choose. An answer engine reads several pages, writes one answer, and names some of its sources. To be one of those sources, a page has to be **easy to lift a passage from**, which is a property of how the page is written, not only of how well it ranks.

## What answer engines need

<Columns cols={2}>
  <Card title="Access" icon="door-open">
    Their crawler has to be allowed to fetch the page. Blocking AI crawlers removes you from cited answers as surely as it removes you from training data.
  </Card>

  <Card title="Structure" icon="list-tree">
    Passages that stand alone. An assistant quotes a section, not your whole article, so a section that only makes sense after reading the previous three is not quotable.
  </Card>

  <Card title="Directness" icon="bullseye">
    The answer near the top. A page that spends four paragraphs on preamble has buried the sentence worth quoting.
  </Card>

  <Card title="Identity" icon="id-card">
    Structured data connecting the page to an author and an organization. Assistants name sources they can identify.
  </Card>
</Columns>

## The retrieval score

Pro adds a retrieval score in the editor, scoring five signals as good, warning, or bad, for a maximum of ten points.

<AccordionGroup>
  <Accordion title="Passage size">
    Words per section, measured between headings. Under 300 is good, up to 500 is a warning, beyond that is bad. A section with no headings at all scores nothing.

    The reason is mechanical: an assistant retrieves chunks. A 1,500-word run of text without a heading is either truncated or skipped.
  </Accordion>

  <Accordion title="Self-contained sections">
    Sections that depend on earlier ones to make sense. A section beginning "This is why it fails" cannot be quoted, because the quote does not say what "it" is.

    Fix by naming the subject again at the start of each section. Slight repetition reads fine and makes each section liftable.
  </Accordion>

  <Accordion title="Question headings">
    Headings phrased as questions. Two or more is good, one is a warning, none is bad.

    People ask assistants questions. A heading that matches the question is the strongest possible signal that the text under it is the answer, and it is what [answer blocks](/rankify/settings/ai-geo) turn into FAQPage markup.
  </Accordion>

  <Accordion title="Answer-first opening">
    Whether the page defines its subject early. Good or bad, no middle.

    Say what the thing is in the first paragraph, then elaborate. The journalistic instinct to build up to a conclusion works against you here.
  </Accordion>

  <Accordion title="Lists and tables">
    Whether the content has any structural formatting. Present is good, absent is a warning.

    Lists and tables are unambiguous to parse and easy to reproduce in an answer. A comparison written as prose is much harder to lift than the same comparison as a table.
  </Accordion>
</AccordionGroup>

<Note>
  The score measures how quotable a page is, not how good it is. A brilliant essay can score badly, and a mediocre FAQ can score well. Treat it as a structural check on pages you want cited, not as an editorial verdict.
</Note>

## What Rankify does for you

<Steps>
  <Step title="Publishes llms.txt">
    A plain-text index of your important pages with summaries, written for models. On by default.
  </Step>

  <Step title="Marks up your answers">
    Pro. Question-shaped headings become FAQPage markup with a concise answer summary, in the format answer engines quote.
  </Step>

  <Step title="Keeps the crawlers allowed">
    AI crawlers are allowed by default, with a per-crawler list if you want to be selective.
  </Step>

  <Step title="Connects your identity">
    The schema graph ties pages to authors and to your organization, and your social profiles confirm it.
  </Step>
</Steps>

## Measuring it

This is the part traditional SEO tools cannot show you.

| Signal                               | What it proves                                                | Where                     |
| ------------------------------------ | ------------------------------------------------------------- | ------------------------- |
| Crawler requests                     | Your pages are reachable                                      | AI crawler log            |
| Distinct crawlers, identity verified | The assistants that matter are reading, and it is really them | AI crawler log, Pro       |
| Most-read pages                      | Which content assistants keep returning to                    | Overview                  |
| Visitors from AI answers             | You were actually cited                                       | AI referral tracking, Pro |

Read it as a funnel. Crawlers reading but no referrals means you are readable but not quotable, which is a content structure problem. No crawlers at all means an access problem, so check robots.txt first.

<Card title="AI and GEO settings" icon="robot" horizontal href="/rankify/settings/ai-geo">
  Every switch on this tab, and what each one does.
</Card>

## Writing a page to be cited

<Steps>
  <Step title="Answer in the first paragraph">
    Then explain. If someone read only the opening, they should have the answer.
  </Step>

  <Step title="Use question headings">
    Phrase them the way a person would ask, not the way a marketer would title a section.
  </Step>

  <Step title="Keep sections under 300 words">
    And make each one make sense alone. Name the subject again rather than relying on "it".
  </Step>

  <Step title="Use tables for comparisons and lists for steps">
    Structure is easier to lift than prose, and it survives being quoted out of context.
  </Step>

  <Step title="Be specific">
    Numbers, versions, dates, and named things. An assistant weighing two sources prefers the one making checkable claims, and vague copy is indistinguishable from every other page on the topic.
  </Step>
</Steps>

<Warning>
  None of this is a substitute for being worth quoting. Answer engines are trying to give a correct answer, and a well-structured page that says nothing specific is still not the one they name.
</Warning>