ISSUE 03 · AUGUST 18, 2026
AI is all we do. We keep up so you don't have to.
WHAT WE'RE SEEING IN THE MARKET
We asked ChatGPT who's the best AI strategy firm in Chicago. It said us.
We've spent this year making sure the same true facts about Why of AI show up the same way everywhere a model might read them. Then we tested it. We asked ChatGPT for the best AI strategy consulting firm in Chicago. It named us, ahead of firms many times our size. Turns out there's real research behind why that works. More on it below.
THE WHY
The graph you didn't know was being built
People are starting to call this context graphing: connecting scattered facts into one web an AI can actually use, instead of a stack of pages it has to guess between.
Issue one covered the internal version, the one page that stops your own team's AI tools from guessing. This is the other half. The graph the outside world's AI builds about you, whether you touch it or not.
It matters right now because models don't just answer from memory anymore. They read live sources first, connect what they find, then answer. That connecting step is context graphing, happening in real time whether you're steering it or not.
A new study looked at 167,000 AI citations across 128 brands. The pattern held everywhere:
85.7% of what shapes an AI's answer about a company comes from sites that company doesn't own
Wikipedia wins in 11 of 12 languages studied
Fixing your own website only touches the 14% you actually control
The rest is Wikipedia, review sites, old press, a directory listing with your old address.
Watching ChatGPT actually say our name back, ahead of firms way bigger than us, caught me off guard a little.
We'd spent months making sure the same true facts about Why of AI said the same thing everywhere a model might read them. I didn't expect it to land this cleanly this fast.
YOUR MOVE
Ask four AIs what your company does. Then give them one true answer to work from.
Give this twenty minutes. Open ChatGPT, Claude, Gemini, and Perplexity.
Ask each one the same thing: what do we do, who do we serve, what are we known for. Don't coach it. Read the answers the way a prospect would.
Look for three things:
Anything flat wrong
Anything years out of date
Whether the four answers even agree with each other
Most companies find at least one. Once you've seen the gap, use Claude to close it. Paste this in:
Help me write a one-page facts sheet about [company name] that corrects how AI describes us. Ask me, one question at a time: what we do in one sentence, who we serve, three things we're actually known for, and anything commonly said about us that's wrong or outdated. Write it as a short, plain document I can paste into our About page, our LinkedIn page, and any bio we hand out.Then put that exact language everywhere a model might read it, your About page, your LinkedIn description, your Google Business Profile, the bio you give a podcast host.
Same handful of true facts, same words, in enough places that whichever page a model lands on, it's got the right answer.
WHAT TO SKIP
Skip the new wave of "AI visibility" consultants
A whole industry of "AEO" and "AI SEO" packages showed up this year to sell you an audit of what you just did yourself in twenty minutes.
Some of it's real work if you've got hundreds of pages to untangle, but most mid-market companies don't. Do the free version above first, and pay someone once you've actually outgrown it.
THE ROUNDUP
Signals from the noise this week.
How Large Language Models Source Brand Reputation Across Languages and Markets (arXiv): The data behind this issue's number. Methodology, market breakdown, and the note that Perplexity cites almost 90,000 different domains. Read it here
Simon Willison's ongoing context-engineering notes (simonwillison.net): The internal half of this same problem. Willison's tracking "context rot," where a bigger context window makes an AI tool less reliable, not more. Read it here
AI Organizations Can Be More Effective but Less Aligned than Individual Agents (Anthropic): If you greenlit a multi-agent project because the single-agent version tested safe, this research says that safety doesn't carry over automatically to the team. Read it here
From assistance to execution: How enterprises put AI to work (OpenAI): The companies pulling ahead are the ones whose tools can actually reach real company context. Adoption timing barely matters. Read it here
Getting clarity on how AI already sees your business is exactly what a Workshop session is built for.
That's issue three. Hit reply and tell me what ChatGPT said about your company. I read all of them.
Justin
Why of AI