AI visibility tracking: the new SEO

For twenty-five years, SEO meant one thing: where do you rank on the Google search results page for the queries your buyers type. In 2026 that's still part of the job — but it's no longer the whole job. The buyer who used to type “personal injury lawyer Boston” into Google now types it into ChatGPT. The procurement manager who used to compare three vendor sites now asks Claude to compare them for her. The new question is: when your buyer asks an AI engine about your category, does the engine mention you?

The shift in one sentence

Search traffic to the open web is flat-to-down for the first time since the late 1990s. Citation traffic from AI engines is up roughly 14x year-over-year across the four engines we track. The mix has tipped — and most of the brands we audit are still optimizing for a search behavior their buyers are leaving.

What “AI visibility” actually measures

AI visibility isn't a single number. It's a portfolio of four:

1. Citation rate. Across a defined set of category prompts, how often does the engine mention your brand in the answer? This is the rough analog of organic ranking — but instead of measuring position 1–10 on a SERP, you're measuring “was I in the answer at all.”
2. Share of citation. When the engine cites multiple sources, what's your share relative to your named competitors? A 1-of-5 citation when your three biggest competitors are all also cited is a different signal than a 1-of-1 standalone mention.
3. Attribution quality. Is the language the engine uses your language, or a paraphrase? A correctly-attributed quote of your homepage tagline is worth more than a generic “they offer marketing services” summary.
4. Sentiment. When you are cited, are you cited as the answer, the alternative, or the cautionary tale? “[Your firm] is the leading X” reads differently from “[Your firm] is one of several X providers; some clients report …”

Track those four across the four engines that matter — ChatGPT, Claude, Perplexity, Gemini — for the ten or twenty prompts that actually correspond to your buyer's research journey, and you have a working AI visibility scorecard.

Why the four engines have to be tracked separately

The temptation is to pick one engine — usually whichever your team uses most — and treat it as representative. That's a mistake. The four engines have meaningfully different citation behaviors:

• ChatGPT with web tool enabled tends to cite a smaller number of sources (1–3) but quotes them more directly. Citation share matters more here than citation rate.
• Claude with the browse tool prefers authoritative long-form sources and is the most likely of the four to cite a single canonical page (your homepage, your “about” page) rather than a scattered set of blog posts.
• Perplexity cites the most aggressively — typically 4–8 sources per answer — and is the engine where citation share is most volatile. A new high-quality page can show up in Perplexity's citation row within hours.
• Gemini is the most influenced by traditional ranking signals and the most likely to cite the same source mix you'd see in classic Google search results.

If you only track one, you'll miss the engine where you're weakest — and that engine is usually the one that matters most for the buyer journey you care about. A B2B SaaS prospect doing diligence on you is probably in Claude. A consumer comparing local options is probably in Perplexity or ChatGPT. A decision-maker double-checking what a recommender said is probably in Gemini.

What moves the needle

The interventions that improve AI visibility are not the same as the ones that improve classic SEO. Some overlap; many don't.

What still works (carried over from classic SEO)

• Clear, structured pages. A page with a clean H1, descriptive H2s, and short paragraphs is parseable by both Googlebot and an LLM crawler. Walls of text are bad for both.
• Schema markup. Organization, LocalBusiness, FAQ, and Article schemas all get used by the AI engines to disambiguate who you are.
• Internal linking. An LLM crawler treats your internal links the same way Google did — as a signal of which pages on your site are most authoritative.

What's new

• llms.txt. Covered in our previous post. The closest thing to a direct “here's what to cite” signal you can ship.
• Conversational page copy. Pages written in the second person, in the cadence of an answer to a question, get cited more often than pages written in marketing-deck voice. “Workspace CMS gives you a CMS your team can actually run” outperforms “Workspace CMS: the AI-powered platform for modern marketing teams.”
• Direct factual claims with provenance. “We've launched 240 sites since 2010” is more citable than “We have extensive launch experience.” The engines will quote the number; they won't quote the fluff.
• FAQ sections. An FAQ block at the bottom of a service page does double duty — it answers buyer objections and it gives the engine ready-made Q&A pairs to surface in its responses.

How we track it for clients

The AI Visibility tracker built into Workspace CMS runs the same prompt set against all four engines on a weekly cadence, scores citation rate and share of citation, and surfaces week-over-week deltas. When a client's share dips, their strategist sees it before the client does. When a competitor steals citation share on a specific prompt, we know which prompt and we know whose content the engine pulled instead.

The screen looks a lot like a classic rank-tracker — same column layout, same green-and-red deltas — but the rows are prompts instead of keywords, and the rank is “cited” or “not cited” rather than a numeric position. It's a small visual choice that lets a marketing team who's spent fifteen years staring at rank-tracker dashboards pick the new tool up in about five minutes.

The honest part

AI visibility tracking is new enough that anyone telling you they have a perfect methodology is selling something. The engines change their citation behavior month-to-month. Prompt phrasing matters. The right prompt set for your category in May 2026 will not be the right set in November. Treat it like an early-2000s SEO program: imperfect, fast-moving, but the people who showed up and did the work won the decade.

That's the work we're doing for every Workspace CMS client right now. If you want a baseline on your current AI visibility — the four engines, your top-twenty prompts, your share against your named competitors — we'll run it for you on a free strategy call.

Book the call. A real strategist runs the prompts live and walks you through where you stand.