The Ghost Citation Problem: When AI Recommends Your Firm Without Saying Your Name

A study published in late July put a number on something immigration attorneys have felt for months without being able to prove it. Analyzing AI answers across the major assistants, researchers found that roughly 40% of the citations AI hands out are "ghost" citations — the engine links to your page as its source, but never says your firm's name in the answer the reader actually sees. On Perplexity, the ghost rate hit 52%. Put plainly: you can be the authority an AI is leaning on to recommend an immigration attorney, and still be invisible to the family reading the answer.

That gap matters more in immigration than in almost any other practice area. A prospective client asking ChatGPT "how do I respond to an RFE on my H-1B?" is not browsing ten blue links. They read one synthesized answer, absorb the guidance, and — if a name is attached to it — remember that name when they decide who to call. When your content powers the answer but your name is stripped out, you've done the work of earning the client's trust and handed the introduction to no one.

What a Ghost Citation Actually Is

AI assistants do two separate things when they use your content. They cite it — a link or footnote pointing back to your page — and they mention it — naming your firm inside the visible text. Attorneys tend to assume the two travel together. They don't. The July research found that citations and named mentions occurred together in only about one in eight appearances. The rest of the time, one happened without the other, and most often it was the citation showing up alone: your URL in the source list, your name nowhere in the sentence a client reads.

For a firm, that's the worst of both worlds. You've produced content strong enough for a model to treat it as ground truth on adjustment of status or consular processing. But the reader who acts on that answer never learns which firm stood behind it. The click-through that used to be your reward for ranking well is now optional, and increasingly skipped — the reader got what they needed inside the answer.

Why It Happens to Good Content

The instinct is to assume ghost citations are a penalty for thin pages. The opposite is often true. Comprehensive, well-structured explainer content is exactly what models pull from to construct an answer — and when a model paraphrases a clear explanation, it frequently keeps the fact and drops the attribution. Your "H-1B RFE response timeline" guide gets absorbed into the answer precisely because it's good, and the better the model's paraphrase, the less it needs to quote you by name.

Two patterns make it worse for immigration firms specifically. First, a lot of firm content is written in the third person about the law rather than about the firm — "USCIS typically issues an RFE within..." — so there's no natural place for the model to anchor your name to the claim. Second, firm names in this space are often long and formal ("The Law Offices of...") and don't read as a natural entity the way a product brand does, so models are less inclined to surface them in prose. The claim survives; the firm doesn't.

The Fix Is Structural, Not More Content

The answer is not to publish more. It's to structure what you publish so your name travels with the specific claim it gets cited for. That means writing so the firm is grammatically attached to the assertion an AI is likely to lift — "At [Firm], we respond to H-1B RFEs within the 87-day window by..." rather than a disembodied statement of the rule. It means making sure the entity the model recognizes — your firm's name, consistent across your site, your profiles, and your author bylines — is the one attached to your most citable claims.

This is where a named-mention gap is measurable. You can look at where AI already cites your firm and check whether your name comes along for the ride. Where it doesn't, the edit is usually small and specific: bind the claim to the firm, tighten the entity signals, and give the model a reason to name you, not just link you.

What to Do This Week

Start narrow, because the fix compounds. A firm doesn't need to overhaul its site — it needs to find the two or three pages doing the most citation work and make sure the name rides along.

1. Pick your five highest-traffic or highest-intent pages — the RFE guide, the H-1B timeline explainer, the green-card-through-marriage walkthrough — the ones most likely to be feeding AI answers.
2. On each, read your own key claims out loud and ask: is my firm's name in the same sentence, or nearby? If the claim floats free of the firm, rewrite it so the two are bound together.
3. Standardize the entity. Use one exact firm name across your site, your Google Business Profile, your directory listings, and your author bios — models reconcile these into a single entity, and inconsistency dilutes the signal.
4. Add a first-person authority line to each cornerstone page — who at the firm handles this work, and a specific, verifiable detail (years practicing, case types, jurisdiction) a model can attach to the recommendation.
5. Re-check in a few weeks. Ask the assistants the questions your clients ask and note whether your name now appears in the answer, not just the source list.

The Bigger Shift

Ghost citations are a preview of where AI search is heading: the reader's attention now lives inside the answer, not on the page it came from. For immigration firms, that raises the stakes on a single question — when an assistant recommends someone for the work you do, is it your name in the sentence? Being the source is no longer enough. The firm that earns the named mention is the one the family remembers when they finally pick up the phone.

Clientory helps immigration law firms become visible to the families and individuals who need them most — turning AI search into consultations, and consultations into clients. See where AI cites you without naming you — book a free visibility review at CLIENTORY.ORG.