Our marketing lead Casey just attended a tech conference and learned how LLMs perceive brands. There's a finding in there about brand names that is hard to believe!
We’ve got some great quality-of-life updates to our products, and a cool use case for GravityBoard.
In this edition:
🤖 The prompt: What AI actually thinks about your brand
✍️ From the blog: Build a self-hosted Canny alternative with GravityBoard
🧰 Updates to the ’Kit
🤖 THE PROMPT
What AI actually thinks about your brand
Casey just attended Ahrefs Evolve Singapore, where Dan Petrovic gave a talk called "I probed Gemini and GPT 1,000,000 times. This is everything I’ve learned so far."
Dan and his team have spent the last year measuring not what LLMs say about brands when you ask them, but the underlying token probabilities driving those answers. A few findings worth sitting with:
The 8× single-token effect. Brand names that tokenize as a single token get recommended roughly eight times more often than multi-token names. GravityKit, we’re afraid, is two tokens 😞 Here’s a tool for calculating your own token count (and they have a good explanation of tokens as well).
The training data is heavily biased. Whole industries and countries are massively underrepresented in how LLMs talk about who does what. If your audience sits in a region the training data underweighted, the model probably doesn’t know you exist…and will happily recommend a US competitor instead.
What moves the needle for LLM optimisation: specificity, quantified outcomes, and verbs over adjectives. "Cuts setup from two hours to five minutes" beats "powerful and intuitive".
Where to start: map your brand’s entity relationships, then strengthen those edges. Don’t chase keywords — chase associations.
If you want to see how Gemini perceives your brand right now, Dan’s team built authority.dejan.ai to make that measurable. Eye-opening if you’ve never thought about it.
If Canny’s per-user pricing has gotten away from you, Casey walks through a self-hosted swap built entirely in WordPress: Gravity Forms collects feature requests, GravityBoard runs a kanban board with native voting, GravityView adds extra data views on top. You keep the data, the URL, and the seat count.
🧰 UPDATES TO THE ’KIT
GravityView 2.60.0 delivered a noticeably faster Views list page on sites with hundreds of forms — we rewrote how the admin screen queries form metadata — and fixed a File Upload regression where uploads were rendering blank on Gravity Forms 2.10+.
DataTables 3.8.0 finally makes per-input filtering possible on Name, Address, and Email-with-Confirmation fields in server-side mode.
Here are all the plugins that got updates since our last newsletter:
We’ve been working on major updates to the GravityView editing experience. Use GravityView? Reply and let me know ONE thing you’d love for us to improve in the View editor.
Until next time Zack with GravityKit
PS - It’s a great time to try GravityView 3.0 beta! We’ve been pleased by how stable it is, and would love for more users to try it before release. If you haven’t given it a spin yet, now’s a great time.
Get this every Friday
One short email a week: new releases, practical Gravity Forms tutorials, and subscriber-only discounts.
