AI models can know a fact about your brand and still fail to retrieve it when the question is asked differently.
New Google research found that frontier models encoded 95% to 98% of tested facts, yet recall remained a major bottleneck.
For SEO, important brand facts should be published from multiple directions.
Fair point. Visibility alone isn’t enough, so the report also measures sentiment and position. I can't name the client for obvious reasons, but the redacted results show positive mentions, a #1.7 average position, and the highest share of voice in the competitive set. So they aren’t simply being mentioned. They’re being presented positively and prominently when users ask relevant commercial questions.
Been deep in AI search visibility lately, and this result was pretty damn cool.
Across hundreds of responses from ChatGPT, Gemini, Google AI, and Perplexity, a client hit:
37.8% AI visibility
29% share of voice
1.7 average position
#1 against their competitors. 😀
Google is turning missed calls into ad spend.
From Oct. 1, a missed LSA call during business hours can become a paid lead if the caller waits 20+ seconds. Some follow-up calls may be billed too.
LSA optimization now extends beyond Google Ads. Your front desk affects your ROI.
I fired my first client today.
I take my clients seriously. The problem is, this client didn’t take the work seriously.
Multiple red flags kept appearing. Deadlines kept getting dragged out. Payments kept getting delayed. Every step became another excuse to buy more time.
At some point, you have to recognize when a client isn’t difficult. They’re just wasting your time.
I work too hard to keep chasing people who have no intention of respecting the agreement.
If you’re going to waste someone’s time, it most definitely won’t be mine.
If you’re going to use LLMs for SEO, at least A/B test between models.
But the model is only part of the equation.
You still need SEO knowledge to know what to ask for, context engineering to give the model the right information, and judgment to know whether the output is actually good.
LLMs don’t replace expertise. They amplify it.
Reddit citations in ChatGPT dropped hard.
Promptwatch found Reddit went from ~3.8% of ChatGPT citations to ~0.5% in a matter of days.
At the same time, ChatGPT appears to have changed how it searches the web, with far more domain-specific site: searches and more fan-out queries per response.
That matters for SEOs.
I break down the data, what changed, and what it could mean for getting cited in ChatGPT.
If a local SEO client trusts you enough to give you access to their personal Facebook account, use it.
Join local Facebook groups where their customers actually spend time: community groups, neighborhood groups, “City + Service” groups, etc.
Then actually participate and post like a real person, not a business page spamming links.
For local businesses, that personal profile + community trust can drive a surprising amount of visibility and leads.
The hardest part?
Your client has to trust you with their personal Facebook account.