Reverse AI Search

Ghost routes: pages that rank on Google but are invisible to AI

Some of your best-ranking pages are ghosts in the answer layer — Google loves them, the AIs never name them. Here's how to spot and revive them.

Updated May 20267 min read
The short answer

A ghost route is a page that ranks well in classic Google search yet is effectively invisible in AI answers — ChatGPT, Gemini and Grok rarely or never cite it for the questions it should win. Ghost routes happen because the two systems reward different things: Google ranks a whole page on links and relevance, while LLMs extract and cite specific, self-contained answer passages. A page can be a #1 organic result and still have no extractable answer block for a model to lift. You find ghost routes by comparing two lists — the queries a page ranks for in search versus the queries the AIs actually cite your domain on (your AI citation footprint). The pages with strong search demand and zero AI citations are your ghosts, and they are usually the cheapest visibility wins you have.

What exactly is a ghost route?

The term borrows from the idea of a route in a sitemap — a real, indexable URL that exists and performs in one channel but is a no-show in another. A ghost route is a URL that:

  • ranks on page one of Google for queries with genuine buyer intent, and
  • is not cited or mentioned by ChatGPT, Gemini or Grok when those same questions are asked.

It is the gap between “findable” and “citable.” A page can be both crawlable and authoritative and still be a ghost in the answer layer, because the AIs are not ranking your page — they are deciding whether a passage of it is the cleanest way to answer the user.

Why do ghost routes happen?

Several mechanics produce them, and most pages that go dark do so for more than one reason:

No extractable answer

Classic SEO pages bury the answer under preamble, or scatter it across the page. LLMs prefer a clean, self-contained passage that answers the literal question in one place. If a model has to synthesise your answer from five paragraphs, it will more often cite a competitor who stated it in one. This is the single most common cause — and the most fixable.

The model leans on a different source type

For some questions the models prefer aggregators, community threads or review platforms over a vendor’s own page. Your page can be correct and well-ranked and still lose to the source the model trusts for that kind of question.

The page targets a query no one asks an AI

Some Google-ranking pages target navigational or long-tail queries that simply don’t get asked of a chat assistant. That is not a page problem — it is a demand mismatch, and it is worth knowing so you don’t waste effort “fixing” a page nobody queries in AI.

Crawler access

If you block AI crawlers (GPTBot, Google-Extended, PerplexityBot and friends) in robots.txt, you can rank in Google while being structurally excluded from the training and retrieval pipelines that feed citations. Check this first — it is a one-line fix.

How do I find my ghost routes?

You need two lists side by side:

  1. Search demand — the queries your pages already rank for (Search Console or your SEO tool of choice).
  2. AI citation footprint — the queries the models actually cite your domain on. Run the free Domain Check to read this across all three models; the concept is covered in AI citation footprint.

Overlay them. Pages with strong search presence and no corresponding AI citations are your ghost routes. This is a close cousin of the AI keyword gap — the gap there is between you and a competitor; the gap here is between your own two channels.

How do I revive a ghost route?

  • Add a self-contained answer block near the top. A 40–60-word direct answer to the page’s core question, phrased so it stands alone if quoted. This is the passage the model can lift.
  • Use the literal question as a heading. One question per H2, answered immediately beneath it, matching how models retrieve.
  • Make facts atomic and attributable. One claim per sentence, with a date and source where relevant. Specific, dated statements are easier to cite than hedged prose.
  • Confirm crawler access. Make sure the page is reachable by AI crawlers.
  • Re-check on a cadence. Reviving a ghost is not instant; AI answers update over time. Monitor the page’s citations so you can confirm the fix landed.

Why ghost routes are the cheapest wins you have

A brand-new topic requires you to earn authority from scratch. A ghost route is the opposite: you have already done the hard work of ranking and earning trust — the page just isn’t packaged for extraction. Repackaging an existing authoritative page is far cheaper than creating demand, which is why ghost routes belong at the top of any reverse AI search worklist. Start by mapping your reverse AI search footprint, then sort by the pages you already win on Google.

Frequently asked questions

Is a ghost route the same as a page that isn't indexed?
No. A ghost route is indexed and ranks well on Google — it is just never cited by ChatGPT, Gemini or Grok on the questions it should win. It is the gap between “findable” and “citable,” not a crawling or indexing failure.
How do I find my ghost routes?
Put two lists side by side: the queries your pages rank for in Google, and your AI citation footprint from a reverse lookup. Pages with strong search demand and zero AI citations are your ghosts. The free Domain Check gives you the second list.
Why does a #1 Google page never get cited by AI?
Usually because there is no extractable answer block for a model to lift, the model trusts a different source type for that question, or AI crawlers are blocked in robots.txt. Google ranks the whole page; an LLM lifts a passage.
How long does it take to revive a ghost route?
Not instantly — AI answers update on their own cadence. After you add an answer block and confirm crawler access, re-check the page’s citations over the following weeks to confirm the fix landed rather than expecting same-day change.