Why AI Describes You Wrong and Not Who You Are
Last updated: July 22, 2026
Maria Dykstra diagnoses why B2B companies are invisible to AI systems.
She built global ad systems at Microsoft that drove $2B in revenue across 1B+ ads per month. She ran TreDigital for 13 years across Fortune 500s and startups. She is the creator of the Algorithmic Authority Stack™ and the Founder Visibility Engine™.
← Guide 1: Identity Fragmentation
AI describes founders by the breakout work that made them visible five years ago, not the work they sell now. The fix is not more content. The fix is diagnosing the dominant association AI is resolving, then building a stronger chain.
Three patterns, all Layer 1.
Six prompts. Four platforms. Five business days.
Re-measure at Day 45 and Day 90.
The four protocols are the starting point of the Founder Visibility Engine.
Your company moved on. AI didn’t.
Ask ChatGPT what a B2B founder is known for. If the answer matches their 2021 specialty rather than the work they sell now, the problem is not simple visibility. When AI describes you wrong it is resolving the wrong version of your entity.
This is the most common preventable positioning failure I see in B2B companies between $5M and $50M. The product evolved. The category shifted. The founder stepped into a new role. AI still describes them as the version of themselves from five years ago.
The cause sits underneath the surface of every AI system. Most founders do not know it exists.
How does AI actually retrieve information about you?
AI systems do not evaluate you like a page trying to rank for a keyword. They assemble an answer from the strongest signals they can resolve about an entity: who it is, what it is associated with, and which associations appear most authoritative.
A knowledge graph is a map of entities and the relationships between them: people, companies, categories, products, methods, publications, and proof. When someone asks an AI system “what does this company do?” or “who is known for this category?”, the answer often reflects the strongest retrievable associations around that entity, not the latest sentence on the homepage.
For most founders, this map looks like a tangled chain:
- “Daniel Okafor” connects to “Acuity Analytics”
- “Acuity Analytics” connects to “open-source observability”
- “open-source observability” connects to “2021 product launch”
- “2021 product launch” connects to “developer tools”
That is the chain AI returns when someone asks about Acuity Analytics today.
Call them Acuity Analytics. I changed the name. I did not change the pattern. Daniel is composited from real B2B audits where the same failure surfaces repeatedly. The product moved on. The founder now sells to mid-market product teams, not developers. The company has thirty-five employees and is preparing for a Series B. The chain doesn’t update.
In the Algorithmic Authority Stack, this is a Layer 1 failure: Identity Fragmentation, caused specifically by Positioning Drift. You changed how you describe yourself. The old versions are still online. AI sees the full historical record, not just the current version.
Failure Pattern · Layer 1 · Positioning Drift
You evolved. AI still describes the old you.
Positioning Drift is the Layer 1 cause pattern in which a company evolved its positioning over time but never updated legacy surfaces. Old press releases, old directory listings, old LinkedIn descriptions still live in the index. All of it is still being read by AI. The company’s current story conflicts with its legacy story in AI’s training data.
Why doesn’t publishing more content fix this?
Because publishing under the wrong dominant association reinforces it.
Acuity’s team produces three pieces of content per week. The CEO writes regular LinkedIn posts. The marketing team publishes case studies. The product team contributes to engineering blogs. Output is not the problem.
Every piece links back to the existing entity chain. The case studies reference “Acuity, the analytics platform built by the team that created Acuity Open Source.” The LinkedIn posts mention “five years of building in observability.” The engineering blogs sit on the original developer-focused subdomain.
The math doesn’t work. New content cannot outrun an old dominant association when the new content still routes back to the old one. The newer positioning gets thirty mentions a month. The older positioning has thousands of indexed references from five years of GitHub stars, Hacker News threads, conference talks, and developer documentation.
This is the Volume Paradox in action: publishing more unstructured content decreases AI visibility, not increases it. Each new post without strong entity signals adds conflicting classification data rather than reinforcing it.
The four diagnostic protocols
Four diagnostic protocols run inside the Algorithmic Authority Stack to surface which Layer is breaking first. They are the starting point of the Founder Visibility Engine, my 90-day implementation system. The protocols run in sequence. Each one depends on the output of the one before it.
Knowledge Graph Audit
The first protocol. Six prompts, four platforms, same day, screenshots and timestamps captured for every response.
The six prompts:
- Who is [Founder Name]?
- What is [Company Name] known for?
- What does [Company Name] do?
- Who are the leading experts in [Category]?
- What’s the best [Service Type] for [Buyer Type]?
- Compare [Company Name] to [Top Competitor].
The same prompts run on OpenAI’s ChatGPT, Perplexity, Google’s Gemini, and Anthropic’s Claude. Responses scored as Strong, Weak, Wrong, or Missing. The output is a documented current state across the four platforms most B2B buyers use.
In the Acuity audit, every response scored Wrong or Missing. All four platforms described the 2021 open-source tool. None of them placed Acuity in any “best analytics platform” recommendation. ChatGPT confused Daniel with a different person of the same name.
That is the diagnostic. No interpretation yet. Just the data.
“I think AI doesn’t really describe us accurately. Maybe ChatGPT says something about the old product? It’s hard to tell.”
“Acuity Analytics is described as an open-source observability tool in 22 of 24 prompt responses across ChatGPT, Perplexity, Gemini, and Claude. Zero placements in real-time analytics platform queries.”
Dominant Association Naming
The second protocol. It runs immediately after the Knowledge Graph Audit. The output is one sentence in this format: “AI describes [client] as [dominant association], not as [intended positioning].”
For Acuity: AI describes Acuity Analytics as an open-source observability tool, not as a real-time analytics platform for mid-market product teams.
That sentence is what makes the program measurable. “Better AI visibility” is not measurable. Nobody knows when it has been achieved. The work spreads across content, SEO, brand, and PR without a clear definition of done.
“Replace open-source observability with real-time analytics platform for mid-market product teams as the dominant association” is measurable. The same six prompts get re-run at Day 45 and Day 90. The screenshot either shows the new association or it does not.
Marketing initiatives keep going. Positioning programs end.
“We want to improve our AI visibility across platforms by producing more thought leadership content aligned with our current positioning.”
“AI describes Acuity Analytics as an open-source observability tool, not as a real-time analytics platform for mid-market product teams. Re-measure Day 45 and Day 90.”
Target Graph Chain Design
The third protocol. The output is a designed sequence in this format:
[Client Name] → [Methodology or Identity] → [Category] → [Buyer Type] → [Outcome Frame] → [Anchor Proof]
Each link has proof. Each link is specific. Each link gives AI a cleaner path to resolve first.
For Acuity Analytics, the target chain looks like:
Daniel Okafor → [Proprietary methodology name] → real-time analytics for mid-market product teams → product leaders preparing for scale → predictable revenue from instrumented growth → $8M ARR with 30+ enterprise customers
The proprietary methodology name is the missing piece. Daniel does not have one yet. Without one, his personal brand collapses into the company brand. The company brand is itself confused. A named methodology creates a new entity in the graph that AI can associate exclusively with him. This is the Semantic Anchor that fixes Layer 3 of the Stack.
This is where most founders get it wrong. They try to kill the old association by removing references to it, rewriting the About page, or pretending the previous work never happened. That doesn’t work.
Mechanism · The Two-Loop Problem
Two AI systems. Two timelines. Two different fixes.
Live retrieval systems (Perplexity AI, ChatGPT Search, Google AI Overviews) update in 2 to 4 weeks when the signal environment changes. Parametric memory (ChatGPT base, Claude, Gemini base models) updates only during model retraining cycles, which run every 6 to 18 months. Fixing one loop while ignoring the other produces the illusion of progress without resolving the underlying entity confusion.
You cannot delete your way out of a stale identity. The fix is not erasure. The fix is to build a new association chain strong enough to win retrieval first.
“Acuity Analytics helps product teams unlock the power of real-time data through innovative analytics solutions that drive business outcomes.”
“Daniel Okafor → [Methodology] → real-time analytics for mid-market product teams → product leaders preparing for scale → predictable revenue from instrumented growth → $8M ARR with 30+ enterprise customers”
Name Collision Detection
The fourth protocol. Every proprietary name candidate goes through a five-step check before it appears on a site, in a bio, or in published content.
- Google search the exact phrase. Wikipedia entries, indexed books, established brands.
- Search the phrase on ChatGPT, Perplexity, Claude. Ask “what is [phrase]?” If a clear entity returns, the name is taken.
- Amazon and Bookshop search. Books are heavy entities.
- Industry context check. Some collisions only matter in the same category.
- Resolution decision. If a collision exists: rephrase, reposition syntactically, or create a distinct shorthand.
Two real collisions surfaced in a single recent engagement.
The first: the client wanted “The Growth Equation” as a category claim. A check surfaced Brad Stulberg’s site at thegrowtheq.com, active since 2019, weekly essays, podcast. Plus Andy Budd’s October 2024 book The Growth Equation, endorsed by Jake Knapp and Jeff Gothelf. Either entity would dominate AI retrieval. The phrase got dropped from the category language entirely.
The second was harder to spot. The methodology umbrella was going to be “The Commercial Operating System.” A search returned Digital Equipment Corporation’s COS-310, a 1970s mainframe with a Wikipedia entry AI returns when asked about the phrase. The resolution was structural rather than substitutive: a distinct shorthand (cOS) became the canonical reference, with the full phrase appearing only on first mention.
Both collisions were caught before the client spent a year publishing into someone else’s entity gravity.
“Our methodology is called The Growth Equation. We’re going to build the whole positioning around it.”
“The Growth Equation collides with Brad Stulberg’s site (active since 2019) and Andy Budd’s October 2024 book. Pick a different name or build into someone else’s entity gravity.”
Founder Visibility Engine™
The four protocols are the first ten days.
The Founder Visibility Engine is the 90-day implementation system for the Algorithmic Authority Stack. The four diagnostic protocols run first. The remaining 80 days execute the fix: identity refresh, schema implementation, content production aligned to the target chain, and re-measurement at Day 45 and Day 90 against the same six prompts.
The protocols surface which Layer is breaking first. The program fixes it.
What three patterns does this audit reveal?
Across the B2B audits I’ve run, three patterns show up again and again. Each one is a different breakdown inside the same Layer 1 of the Stack.
Positioning Drift
AI describes the founder by the breakout work that made them visible five years ago, not the work they sell now. You changed how you describe yourself. The old versions are still online. AI sees the full historical record, not just the current version.
This is the Acuity Analytics pattern. Most common in founders whose product or category has evolved faster than their owned surfaces.
Identity Fragmentation through adjacent entities
The founder and the company are two entities, but AI cannot resolve them as separate. The founder gets described by the company’s specialty even after stepping into a higher-value category, product, or advisory role.
Agency founders who have moved into independent practice, fractional roles, or advisory work see this constantly.
Authority Collapse
The founder may be excellent. The public entity is weak. Knowledge exists. The structural proof signals AI needs to verify expertise do not.
This shows up most often in founders who built their reputation inside larger organizations and have only recently established a public footprint of their own.
Three patterns · One first step
Different failure. Different fix. Same first step: diagnose the dominant association before producing more content.
What does this look like in practice?
If you suspect AI is resolving the wrong version of you, the four diagnostic protocols can be run in five business days.
I built a lighter version of this diagnostic for founders who want to know which problem they actually have: Positioning Drift, Identity Fragmentation through adjacent entities, Authority Collapse, or something else.
The AI Visibility Snapshot is the lighter version of this diagnostic. You get a written current-state diagnosis, three structural gaps, and the exact pattern AI appears to be resolving around you. No strategy deck. No sales theater. Diagnosis first.
If the diagnosis surfaces work worth doing, the full Knowledge Graph Audit adds the Target Graph Chain Design, Name Collision Detection on every proprietary term, and the 90-day execution plan inside the Founder Visibility Engine.
If the diagnosis surfaces nothing, you have lost twenty minutes.
FAQ
Why does AI describe me by my old positioning instead of my current work?
AI systems retrieve entities by their strongest associations, not by the latest sentence on your homepage. If your breakout work from five years ago has thousands of indexed references and your current positioning has thirty, AI returns the older association.
This is Positioning Drift, a Layer 1 failure in the Algorithmic Authority Stack. The fix is not erasure. The fix is building a new association chain stronger than the old one.
What is the Knowledge Graph Audit?
The first of four diagnostic protocols inside the Algorithmic Authority Stack. Six standardized prompts run across four AI platforms (OpenAI’s ChatGPT, Perplexity, Google’s Gemini, Anthropic’s Claude) on the same day, with screenshots and timestamps captured for every response.
The output is a documented current state. No interpretation. Just the data. It is the starting point of the Founder Visibility Engine.
What is the Two-Loop Problem?
The structural AI visibility failure that occurs after a rebrand or acquisition. The live retrieval layer (Perplexity AI, ChatGPT Search, Google AI Overviews) and the parametric memory layer (ChatGPT base, Claude, Gemini base models) require entirely different fixes on entirely different timelines.
Live retrieval fixes in 2 to 4 weeks. Parametric memory fixes in 6 to 18 months. Fixing one loop while ignoring the other produces the illusion of progress.
How do I know if my positioning has drifted in the AI knowledge graph?
Run the six diagnostic prompts on ChatGPT, Perplexity, Gemini, and Claude. Ask who you are, what your company is known for, what your company does, who the leading experts in your category are, what the best provider for your buyer type is, and how you compare to your top competitor.
If any platform describes you by work you no longer do, you have Positioning Drift. The AI Visibility Snapshot formalizes this diagnostic across four platforms.
Why doesn’t publishing more content fix the wrong AI association?
New content cannot outrun an old dominant association when the new content still routes back to the old one. If every blog post, LinkedIn update, and case study references the legacy positioning, more content strengthens the wrong association rather than replacing it.
This is the Volume Paradox. Publishing under the wrong dominant association reinforces it instead of fixing it.
What does a Name Collision check actually do?
Name Collision Detection prevents proprietary names from colliding with existing entities in the AI knowledge graph. Every methodology name, framework name, signature phrase, or category claim goes through a five-step check before locking.
Two recent collisions surfaced in a single engagement: The Growth Equation (Brad Stulberg’s site and Andy Budd’s October 2024 book) and Commercial Operating System (Digital Equipment Corporation’s COS-310, 1970s mainframe with a Wikipedia entry). Both were caught before the client spent a year publishing into someone else’s entity gravity.
Related reading:
- Guide 1: Identity Fragmentation
- The Algorithmic Authority Stack: 7 Layers Between You and AI Visibility
- Layer 1: Market Identity Clarity
- The Founder Visibility Engine: 90-Day Implementation System
- How to Fix AI Visibility After a Rebrand or Acquisition
Is your company invisible to AI?
Six questions, about 90 seconds. Find out which of the seven layers is breaking first, and whether you are failing to be retrieved or failing to be cited.
Take the AI Visibility Test
