What kind of website gets cited by every AI engine?

Zero. That is how many of our 29 domains cleared the bar we set ourselves.

Seven days earlier, the same tool, pointed at a different set of Indian page-one domains, said 23 of 28 cleared it. Two domains from that earlier run were re-measured today, and what happened to them is the whole story.

Zero of 29, and what changed in seven days

Method, fixed in writing before the run:

• 1

  Pull the Google India organic top 10 for four commercial buyer queries, one each in four unrelated industries

• 2

  Reduce every result to its registrable domain and dedupe across queries, giving 29

• 3

  Ask an AI citation index how often each of seven engines cited each domain

Six engines returned numbers. Grok returned 0 for all 29 domains, including youtube.com, which the same call reports at 961,792 ChatGPT citations, and reddit.com at 3,485,711. A universal zero across a set spanning six orders of magnitude of citation volume is not engine behaviour.

How do you tell an index gap from a real absence?

You re-measure something you already have a reading for. We took two domains from our 14 August run, both with non-zero Grok counts then, and called the identical tool with identical settings.

Perplexity, Gemini, AI Overviews, AI Mode and Copilot moved by a few percent at most on both domains across those seven days. That is a controlled before and after on a fixed pair. It rules out our own settings and points at the index. We measured the index and not Grok, so this says nothing about whether Grok itself cites these sites.

Is cited by every AI engine worth anything?

Drop Grok and the picture inverts completely. All 29 of 29 domains are cited by all six surviving engines. Not most. Every one.

That sounds like good news until you look at volume.

Nineteen citations and twenty five million both earn the same badge. The two ends of this set are separated by a factor of roughly 1.34 million, and the metric cannot tell them apart.

"Nineteen citations and twenty five million both count as cited by every engine. The badge is a threshold, and the threshold is one."

We also chose that threshold. Cited by all seven engines is our line, not a definition any of these platforms publishes. Our older reading of what the engines disagree about used the same line, which is why its 82.1%, or 23 of 28 domains, and this zero do not really conflict.

What actually tracks how often you get cited?

Since presence did not separate anything, we looked at volume instead. This part was not planned before the run, so treat it as a direction rather than a result.

Across the 29 domains, rank correlation with six-engine citation totals came out like this:

How many queries you already rank for tracks citation volume more tightly than how strong your backlink profile is. We found the same ordering on 27 Coimbatore businesses three days earlier, on a completely different population. Two samples agreeing is worth noting. It is not proof, and it does not show that adding keywords causes citations.

What does Google say about tools like this one?

Bluntly, that they cannot see inside. Google guidance on third-party SEO tools is unusually direct about the limits of exactly the class of instrument we just used.

"Third-party tools don't have access to our internal ranking data. They can't guarantee performance. Any predictions are their own and like predictions generally, may not happen."

Its guidance on optimizing for generative AI search puts it more sharply still, and also names the first-party alternative for the Google surfaces. Two things in that guidance are worth separating:

• Google is describing its own systems, so the warning is authoritative about Google and about nothing else
• It has no view into how ChatGPT or Perplexity pick sources, so the caution carries over for a different reason: nobody outside those companies can see in either

How do you check AI visibility without trusting one index?

Use the first-party sources where they exist and treat everything else as a signal with an error bar. Google is explicit about which one it wants you to use.

"Be wary of third-party tools that promise ranking success or claim to use internal Google metrics. No third-party tool has access to our internal ranking or AI systems."

Two of the engines publish something you can act on directly. Our walkthrough of the two failure modes covers the supply side in more depth.

• 1

  Read the Search Console generative AI report for the Google surfaces, which is Google own data rather than a reconstruction

• 2

  Allow PerplexityBot in robots.txt and permit its published IP ranges through your firewall

• 3

  Ask any AI visibility vendor which engines they cover and when that coverage last changed

• 4

  Track any citation count as a quarterly trend, never as a monthly number

Perplexity crawler documentation states that PerplexityBot surfaces and links websites in its search results and is not used to crawl content for AI foundation models, and recommends allowing it. That is concrete and checkable, unlike a dashboard number you cannot audit.

What this does not show

Four queries, one country, one day, against one prior day. Two readings are a change, not a trend.

• Every domain here already ranks in the Google India top 10, so this says nothing about sites that do not rank
• This is an index reading, not our own prompting, so a zero can mean the engine skips the site or that the index cannot see it
• Entry criterion was the organic top 10 for four commercial queries
• The correlation analysis was added after the planned comparison failed
• One of the four queries shares a sector with a query we used on 17 August
• Gemini carries no headline here, because its coverage in this index has been visibly thin since 14 August

If you want the same measurement run against your own domain set with the caveats written down rather than hidden, that is what our AI visibility work is.