1. GEO is your brand on a screen in front of a buyer, so measure it like media, not SEO.
What's happening: If you'll pay per impression for a YouTube pre-roll, an LLM recommendation isn't worth nothing. Wil and Alisa shared six KPIs leadership already understands: share of voice and AI shortlist rate (be seen), message pull-through and brand accuracy (be believed), and pipeline contribution and recommendation rate (be chosen).
2. Being right about you isn't the same as citing you.
What's happening: Alisa ran a brand canon check on a high-profile 2026 Senate race and ChatGPT matched the official campaign positions and press coverage 100% of the time. But it never cited the candidate's own page on data centers, because that position lived on a catch-all issues page instead of a page built to answer the question.
3. A mention in your tracker can be a warning, not a win.
What's happening: In a CRM buying prompt, one brand was named but framed as "I would skip this" because its AI was "bolted on." Mention-only tracking logs that as visibility. Meanwhile, models are leaning less on generic "best CRM" searches in query fan-out (thanks to listicle spam) and more on searching specific brand names, so the brands hardwired to a category get shortlisted before the answer is written.
4. One review can steer the answer, and your own site may not be in the room.
What's happening: When the conversation moved to trust, ChatGPT pulled a single Trustpilot reviewer into its response about a brand with 16,000 reviews on G2 alone, and none of the cited sources were the brand's own site. Wil has seen one negative review affect 40% of his tracked prompts.
Action Items
1. Write your brand canon. Have AI pull your facts from official sources, then fix what it gets wrong at the source. What would it get wrong about you today?
2. Count how often AI cites your own site for the prompts that matter, and give your biggest questions their own page. If you announce it in a press release, is there a page on your site that says it too?
3. Check your highest-stakes facts daily, like pricing, eligibility, and legal claims. Would you know if AI got one wrong before a customer did?
.png)
-png.png)
.png)
.png)
