There is a mechanical reason some pages get quoted in AI answers and near-identical pages do not, and Drew Dorenfest of Client Magnet CRM described it on the Unscripted SEO podcast in one line: the sources you name are the sources a crawler can check.
The Verification Loop Behind AI Citations
“The key to getting cited is to cite your sources.”
Drew Dorenfest, Client Magnet CRM
The loop has four steps. You publish a claim. You name where the claim came from. A crawler follows the named source and compares it against what you wrote. If it matches, the page earns a trust signal that is far harder to fake than a link.
“When crawlers come and cross-reference your information against those sources and verify them, they go: okay, this company is legitimate, has been around for ten years, has great reviews — I will treat this as an authoritative source.”
Drew Dorenfest
Note what is being verified. Not your prose quality. The correspondence between your claim and a checkable external record.
What Dorenfest Means By Pointing At The Source
The instinct most content teams have is to keep the reader on the page. Dorenfest’s argument runs the other way: an outbound link to a primary record is the cheapest credibility you can buy, and it costs you a reader far less often than the traffic-hoarding instinct assumes.
“Linking outward to IRS.gov and other authoritative sources signals verifiability to crawlers and LLMs.”
Drew Dorenfest
Government records, standards bodies, regulator filings, published research, first-party data with a methodology attached. These are the citations that survive a check. A link to another blog post that also has no source is a dead end for the crawler and does nothing for you.
Citation Formats That Survive Parsing
Three formats, in ascending order of usefulness:
- The bare hyperlink. A word or two of anchor text pointing somewhere. Weakest — nothing in the sentence tells a parser what is being asserted.
- The named-source sentence. The source is named in the prose, in text, alongside the link. Now the claim and the attribution sit in the same sentence.
- The fully attributed statistic. Figure, unit, period, named source, link, and date of retrieval. This is the one that gets lifted whole into an answer.
The third format is more work per claim and it is why most pages never do it. That is also exactly why it differentiates.
Auditing An Existing Article For Missing Sources
Take your ten highest-intent pages. Read each one with a highlighter on every sentence that states a fact, a number, a date or a comparison. Every highlighted sentence is a citation slot.
Most teams find the same pattern: the introduction is full of unsourced industry claims, the body is fine, and the conclusion re-asserts the unsourced claims. Fix the introduction first — it is the part most likely to be excerpted.
The step-by-step version of this pass is written up as an SOP in get cited by ChatGPT, and the underlying principle sits in the strategy library as cite to get cited.
The Compounding Effect On Entity Trust
One well-sourced page is a page. Forty of them, consistently formatted, across a single topic, is an entity that models have repeatedly found to be checkable. That is a different asset, and it is the one that keeps paying.
This is where source discipline meets entity engineering for AI visibility: the citations are what make the entity legible, and the entity is what makes the citations compound.