If you've searched your own agency in ChatGPT or Perplexity and got nothing back, don't panic. You're in the majority, not the exception. But there's almost always a specific, fixable reason for it. In nearly every case it comes down to one or more of three gaps: no consistent positioning, nothing structured for a model to lift cleanly, or no repeated signal across enough content. Here's how to tell which ones apply to you.

Invisibility is the default

Boring Marketing's 2026 audit of real buyer questions found 50% of the brands checked were invisible on all four major platforms - ChatGPT, Perplexity, Claude and Gemini - at once. Of brands that do get cited, visibility concentrates heavily at the top: an analysis of 58.6 million AI citations found the top 15 domains across the web capture roughly 68% of all citation share, a tighter concentration than organic Google results ever showed.

So the starting point is straightforward: invisibility is the default state. Something deliberate must change to shift that.

Reason 1: no consistent positioning

Most agency websites use interchangeable language: "full-service," "creative-led," "results-driven partner." This isn't a tone issue; it's a retrieval problem. Models selecting sources need clear, specific, repeatable claims they can confidently cite. Vague messaging provides nothing to extract.

Princeton and IIT Delhi's GEO (Generative Engine Optimization) study, presented at KDD 2024, found that adding quotations from credible sources lifted a page's visibility in generative answers by up to 40%, citing sources by 27%, and adding specific statistics by 34%. Specific, confident claims get amplified. Hedged, generic ones disappear.

If your homepage is indistinguishable from competitors', models experience the same confusion.

Reason 2: nothing structured for extraction

Even precise claims become invisible if models can't access them. Common barriers include case studies as PDF downloads, results in JavaScript-rendered pages crawlers can't parse, and pages lacking clear heading structure with direct answers positioned prominently.

FAQ schema markup's effectiveness has shifted, too. Google removed FAQ rich results from search entirely on 7 May 2026, and testing since shows the markup itself produces little measurable impact on citation rates. What matters is visible, readable Q&A content - all major models extract this with or without underlying markup.

Platform behaviour differs significantly. Separate 2026 studies put Perplexity's brand citation rate at roughly 13-14% on average, against ChatGPT's under 1% - a gap of well over ten times. Plain, clean HTML with direct answers positioned prominently remains the closest shared standard across platforms.

Reason 3: no repeated signal

This catches most agencies because it's least intuitive. A single strong page, even specific and well-structured, typically isn't sufficient alone. Models build trust like people do: by encountering consistent claims repeated across multiple sources that don't contradict themselves.

My own case study is a plain example of this: a UK digital marketing agency with zero AI search presence, cited by all 8 major LLMs within eight weeks - not from constant publishing, but from six content pieces reinforcing identical positioning on the same commercial prompts. Visibility climbed from 0% to 30%, generating 2,357 citations across all eight platforms in a single week.

Six articles. Repeated, consistent claims until patterns became established fact rather than passing mentions.

The fix: define the pillars first, then repeat them

Vague positioning, unextractable content, and isolated mentions appear separate. They're not. All three stem from missing one discipline: the Authority Framework. Identify the small number of things your agency is genuinely, specifically recognised for, then publish focused content reinforcing those same claims on defined prompts, structured for clean model extraction. Maintain consistency rather than sporadic effort.

Practically:

  • Identify three or four specific, truthful claims about your agency before writing. Not five. Not a dozen.
  • Structure every piece with direct answers near the top, clear headings, and plain HTML. No PDFs, no unrendered JavaScript walls.
  • Publish small, consistent content addressing those same claims and target prompts, rather than scattered calendars.
  • Review actual results every few weeks and adjust. Visibility here is measurable; guesswork becomes unnecessary quickly.

Most of this isn't complex. It's just not executed consistently enough for models to register.

A few common follow-up questions

Is this the same as SEO?

No. Related discipline, different mechanism. SEO earns ranking positions. This earns citations within generated answers. Both are worth the effort and reinforce each other, but neither replaces the other.

How long does this take to show results?

My case study above showed measurable visibility across all 8 major LLMs within eight weeks from a zero starting point. Results depend on the existing groundwork, but this represents a realistic timeframe.

Do I need a content team to do this?

No. The case above involved six content pieces, not sixty. The constraint is consistency and clarity, not team size.