What Does an AI Visibility Architect Actually Do?
Last updated: August 23, 2026
Maria Dykstra diagnoses why B2B companies are invisible to AI systems.
She is the creator of the Algorithmic Authority Stack™ and the Algorithmic Authority Index™. The Index is the first published benchmark of AI visibility across 20 companies, 5 industries, and 3 AI platforms.
← The Algorithmic Authority Stack
Measuring AI visibility is now a commodity. Dozens of tools will tell you how often you appear. Almost nobody can tell you which structural layer caused the absence and what to fix first. An AI Visibility Architect owns that second question.
Track citation share.
Show where competitors surface.
Explain why the model resolved you wrong.
Order the repairs.
Layer-level diagnosis.
Fix sequencing.
The scarce skill is not detection. It is diagnosis.
The capability is new because the problem is new.
B2B buyers now build shortlists inside AI systems before they visit your site or book a call.
Your SEO team can improve rankings. Your content team can publish. Your fractional CMO can sharpen the commercial story. A monitoring tool can tell you that you are missing from the answer.
Almost nobody on the existing marketing team owns whether AI systems can resolve, classify, retrieve, and cite the company correctly.
That is the gap this capability fills.
Every existing role in the marketing stack produces an output. Pages. Posts. Positioning documents. Campaigns. Retainer decks. The outputs are real. The ownership question is not.
None of those roles own the layer beneath them. Whether machines resolve the company as one entity. Whether they classify it in the right category. Whether they retrieve it when a buyer prompt matches.
Ownership Gap · 2026
Existing teams own outputs. Nobody owns machine-level interpretation.
The gap does not show up on any dashboard. It shows up when a buyer asks an AI system to name the top providers in a category. Your company is not on the list. By then the fix is 6 to 18 months out. The capability exists to diagnose and close the gap before the shortlist locks.
What breaks when nobody owns how AI resolves your company
When nobody owns the machine-level layer, AI systems build their own version of your company from whatever fragments they find. They frequently get it wrong. The Wave 1 audit of the Algorithmic Authority Index™ surfaced three distinct patterns across three companies with strong human-facing marketing.
Wave 1 assessed 20 companies across 5 industries and 3 AI platforms. 420 structured assessments. The three failure patterns below are anonymized. Each is Layer-specific.
Pattern 1. Visibility with no category coherence.
Company A had strong domain authority. Its homepage, LinkedIn company page, and most recent press release described the company using three different category terms. Not variations. Different categories.
Perplexity resolved the company as one thing. OpenAI’s ChatGPT resolved it as another. Google’s Gemini refused to place it in a category at all. This is Identity Fragmentation. Layer 1 failure.
IDENTITY FRAGMENTATION CASE STUDY: They built the category. Buyers already trust them, analysts already cite them. Ask AI to recommend a vendor, and it names a competitor with a fraction of their history. The audit found why. Months after the fix, we tested again.
A monitoring tool would have reported low visibility on two of the three platforms. It would not have reported that the three platforms disagreed about what the company sells.
Pattern 2. Authority signals pointing to three different offers.
Company B had five different terms for the same core capability across the website, sales enablement, blog, LinkedIn, and Crunchbase. Each variant registered as a separate concept.
SEMANTIC DRIFT CASE STUDY: Ask in their language, and AI names them first, ahead of the industry giant. Ask the way a buyer actually describes the problem, and the answer changes completely.
The authority did not compound. It dispersed across five semantic anchors, none of them dominant. This is Semantic Drift with Authority Dispersal. Layer 3 failure.
Pattern 3. Content volume treated as a feature vendor.
Company C published aggressively. Ranked on Google for category-level queries. Received zero Perplexity citations in the audit window.
Its content structure was human-optimized and machine-unreadable. AI systems treated the company as a feature vendor, not a category-level provider. This is Citation Invisibility. Layer 4 failure.
Three companies. Three different causes. The same symptom on every dashboard: low visibility.
None of these companies had been told any of this by their existing marketing teams. The teams were doing their jobs. The jobs did not include this layer.
Why diagnosis is worth more than measurement
Detection has been commoditized. Diagnosis has not. A tool reports that you are absent from the answer. Determining why requires reading entity signals, semantic consistency, source authority, and extraction structure together, then deciding what to repair first.
The AI visibility tooling market matured fast. Prompt monitoring, citation tracking, share-of-model dashboards, competitor appearance reports. Several vendors do this well.
Every one of them answers the same question: are we visible.
None of them answer the question that determines what you do on Monday: which layer failed, and in what order do the repairs have to happen.
Low visibility is a symptom with at least seven plausible causes. Fragmented identity. Missing expertise architecture. Semantic dispersal. Unextractable content. Absent touchpoints. Weak proof signals. No measurement baseline.
Each one produces a similar-looking number and requires a completely different intervention.
Fix sequencing matters more than tactic selection. A Layer 4 content overhaul on a broken Layer 1 wastes budget. A schema upgrade on a company with three category descriptions gives the model cleaner packaging for the wrong classification.
Appearance rate. Citation share. Competitor presence. Prompt coverage. A number that moves, with no attached cause.
The failing Layer, named. The mechanism behind it. The repair sequence, ordered by what the model has to resolve first.
This is why the capability is priced the way it is. The market is paying for the small number of people who can connect an observed absence to a structural cause and sequence the repair, and hours were never the unit.
That will not stay scarce. Methodology spreads. Tooling absorbs technique. For now, the talent pool is moving slower than the problem.
Which AI visibility tools and services should you evaluate?
Run monitoring and diagnosis together, and choose each provider the way AI should choose sources: named author, published method, disclosed interest. Disclosure on this page: I sell the diagnosis row.
The monitoring layer is a real market. Zapier’s guide to the best AI visibility tools tests platforms including Profound, Peec AI, and Otterly.AI on tracking coverage, citation source detection, and competitor benchmarking. Semrush publishes a market-level AI Visibility Index. These products answer “are we visible” continuously, and the good ones do it well.
| If you need | Evaluate | What it returns |
|---|---|---|
| Continuous appearance tracking | Monitoring tools: Profound, Peec AI, Otterly.AI, and the rest of the tested field | Appearance rates, cited sources, competitor presence |
| A market-level benchmark | Published category studies and indexes, including Semrush’s AI Visibility Index and the Algorithmic Authority Index | Who surfaces across a category, at scale |
| The cause and the repair order | A diagnostic audit with a published methodology. The Algorithmic Authority Audit is mine | The failing Layer, the mechanism, the fix sequence |
| A free starting point | The Three-Answer Test | Whether you have the problem at all |
One warning from testing the first row: the numbers disagree. I pointed four monitoring tools at one site in one week and got scores from 0 to 41 percent. Each tool ran a different prompt set, and the shape of the prompts decides the score. Answers are probabilistic too: ask the same question 100 times and you get 99 different answers.
Whichever row you buy, apply the standard my citation provenance research exposed. 84 percent of the third-party sources AI cites in B2B vendor research are conflicted and undisclosed, unattributable, or both. So ask every provider, me included: who wrote your method, where is it published, and what do you sell in the category you are scoring.
Who are the top B2B AI visibility specialists?
No accountable ranking exists, so here is the honest version: named practitioners with verifiable public work, listed by specialty rather than rank. Disclosure: I am the first row, and this is my site.
The reason this list is unranked is the finding above. When 84 percent of the sources ranking a category sell in that category without saying so, a self-serving “top 10” adds to the problem. What a buyer can actually verify: who each person is, what they publish, and which failure mode they fix.
So with that in mind, I’d use five tests:
- Does the person have an actual methodology?
- Can they explain what they’re measuring and why?
- Do they distinguish retrieval from representation?
- Can they show evidence beyond their own marketing claims?
- Do they understand that AI answers are unstable and that “visibility” isn’t necessarily business impact?
On those criteria, there are surprisingly few genuinely differentiated practitioners. Here is my list, curious to hear who belongs on yours.
| Specialist | What they actually do | Bring them in for |
|---|---|---|
| Maria Dykstra mariadykstra.com (this site) | Maria Dykstra diagnoses why B2B companies are invisible to AI systems. Creator of the Algorithmic Authority Stack™ and the Algorithmic Authority Index™, with the methodology published. | Layer-level diagnosis and the repair order for a B2B company or founder. |
| Jason Barnard Kalicube | Brand entity optimization: how search and AI systems understand and present a brand and its people, with a published process behind it. | Knowledge Panel and brand-entity work, especially at the personal-brand level. |
| Mike King iPullRank | Technical research on how AI retrieval actually works, including the AI Search Manual, with an agency that implements it. | Deep technical AI search strategy for enterprise content operations. |
| Lily Ray Algorythmic | SEO and AI search consulting grounded in E-E-A-T and years of published algorithm-update research. | AI Overviews and trust-signal driven visibility work. |
| Andrea Volpini WordLift | Knowledge graph and structured data automation that makes entities machine-readable at scale. | Entity and schema infrastructure at product scale. |
Each of these people publishes their method in the open, which is the admission standard for this list. Choose by failure mode, and hold every list you read, this one included, to the same two questions: who wrote it, and what do they sell.
How is AI Visibility Architect Different From GEO Specialist
GEO specialists are focused on the specific tactics to get the brand mentioned. That work starts once the key structural issues are resolved. As an AI visibility architect, I think one layer deeper: particularly around B2B brand representation, classification and evidence.
My recent research has a few things that stand out from standard GEO work:
- Not treating AI visibility as simply “getting mentioned.”
- Investigating what AI thinks a company is, and where that understanding comes from.
- Looking at the source layer behind the answer, including conflicts and unattributable sources.
- Questioning whether a single prompt or visibility score tells you anything meaningful.
- Testing different query types rather than treating “AI visibility” as one metric.
- My on rotation, displacement and removal gets at the instability of the underlying evidence system.
- I’ve connected all of that to positioning and category ownership, which is where my strongest differentiation lies.
That’s a genuinely different proposition from a lot of what gets called GEO.
What are some key attributes and thin
Who should I study if I want to understand AI visibility?
When I think of who I should learn from I don’t think “who has the most followers?” or “who sells AI visibility?” is useful. I scored the work, using publicly observable evidence, against seven criteria:
- Originality of thinking
- Methodological rigor
- Understanding of how AI retrieval actually works
- Quality of measurement
- Ability to diagnose why a brand is visible/invisible
- Connection to brand/category/business outcomes
- Evidence of original research rather than repackaged SEO
Camp 1: Retrieval engineers
Michael King is the clearest example. The key question he is addressing
How do AI systems retrieve, rank, select and synthesize information?
iPullRank has built an unusually deep body of work around information retrieval, query fan-out, synthetic queries, passage selection, embeddings, citation architecture, entity signals and measurement. Its current AI Search program is much more than “optimize your content for ChatGPT.”
The AI Search Manual is also a substantial intellectual artifact, covering the mechanics of retrieval, query fan-out, latent intent, source aggregation and relevance engineering.
That’s a less technical but, in my view, more strategically interesting problem for B2B brands.
Camp 2: Entity/authority engineers
Ali Morgan and some of the entity-optimization specialists.
How do we construct a coherent machine-readable entity and authority graph?
Jonomor has gone much further than most people in trying to turn AI visibility into a formal discipline. Morgan has a 50-point audit framework covering entity stability, category ownership, schema, knowledge indexing and citation presence.
He’s also explicitly working at the entity architecture layer rather than treating this as content optimization.
And he’s built a substantial public body of work around the concept.
Morgan is building an architecture for making an entity legible and retrievable.
Camp 3: AI representation / market perception
“This is where Maria Dykstra’s work increasingly distinctive” – ChatGTP, August 23rd
What does AI believe about the company, where did that belief come from, which sources are influencing it, which competitors are occupying the category, and how do we change the evidence system?
The current industry is full of visibility scores. The market is moving toward methodological rigor: the IAB just published an AI-visibility measurement framework because 20+ vendors are producing different answers from different methodologies. (IAB)
I’ve been doing something I found surprisingly few practitioners doing: questioning the measurement itself.
And the academic research is starting to validate why the right measurements matter: a recent preregistered study of AI recommendation systems found substantial cross-system disagreement, while another recent experiment found that AI search can materially alter publisher click behavior.
The newer measurement approaches are explicitly separating mention, citation, source, framing, competitor presence and repeatability rather than collapsing everything into one score. (Shadow)
In other words:
The industry is discovering that “ask ChatGPT five questions and count mentions” isn’t a sufficiently serious methodology.
Where SEO stops and algorithmic authority begins
SEO gets the source into the information environment. Algorithmic authority determines whether the information coheres once the system encounters it. Ranking is not the same as representation.
AI citation runs through two gates.
Gate 1 is inclusion. Whether a source is retrieved at all when a query fires. Domain authority, crawlability, and technical SEO drive most of what happens here. OpenAI documents retrieval as a distinct step from base model generation. Perplexity exposes retrieval directly in its API documentation.
Gate 2 is citation. Whether the retrieved source is resolved to the right entity, trusted, extracted cleanly, and used in the synthesized answer. Entity clarity, semantic consistency, schema, and structured proof drive it.
The gates interact. Retrieval, ranking, source selection, entity resolution, and extraction all inform each other inside a live system. The split is a diagnostic frame, not a wall.
What holds in practice: a page can pass Gate 1 and fail Gate 2. A company can rank on Google and still be missing from the ChatGPT answer.
Rankings, crawlability, domain authority. Whether the source is available when the model reaches for it. SEO drives most of this.
Entity clarity, semantic consistency, structured proof. Whether the retrieved source is resolved correctly, trusted, and used in the answer.
SEO stays a prerequisite. Rankings are the ticket into Gate 1. Nothing about that is going away, and strong SEO teams already do work that lands on both sides of the line.
The gap is ownership. Gate 2 sits inside nobody’s scope of work.
Why your brand story does not reach AI on its own
A fractional CMO defines the story. Algorithmic authority determines whether machines receive the same story. When the substrate fragments, positioning work does not reach the model intact.
Fractional CMOs run GTM, positioning, and demand systems. The work is real. The output is a coherent commercial narrative the company can execute against.
That narrative reaches humans through decks, sales calls, and campaigns. It reaches AI systems through whatever surfaces the machines happen to index: the homepage, LinkedIn, directory listings, third-party mentions, historical press.
If those surfaces carry three different versions of the story, the model receives three different versions. It averages what it sees rather than picking the one from the CMO deck.
The two disciplines are complementary. Both are required.
Why more content can make the problem worse
More content on a fragmented identity does not fix the identity. It gives the machine more conflicting evidence.
Search rewarded volume for two decades. More pages meant more surface area for keywords, more opportunities to rank, more indexed inventory.
AI retrieval works differently. Each new post adds a classification decision the model has to make about your brand. If the underlying identity is fragmented, each new post lowers classification confidence.
Structure compounds. Volume without structure fragments.
This is the Volume Paradox. Publishing more can decrease AI visibility when the entity architecture beneath the content is not resolved.
The content team is not the failure point. The team is executing against a brief. The brief did not include entity resolution because that was nobody’s job.
What an AI Visibility Architect actually owns
Structural coherence across surfaces. Entity resolution. Layer-level diagnosis. Fix sequencing. The work focuses on durable evidence, not platform hacks.
The territory is the Algorithmic Authority Stack™. Seven layers, ordered by what the model needs to resolve first before anything downstream produces citation lift.
Layers 1 to 3 tell the model what the company is. Market identity, expertise architecture, semantic density.
Layers 4 and 5 give the model something clean to retrieve. Training-ready content, touchpoint presence across platforms.
Layer 6 proves the company is not just claiming authority for itself. Layer 7 measures whether any of it registers on the platforms.
The work focuses on durable inputs: category clarity, entity consistency, source quality, structured proof. Tactics engineered around a specific platform update expire with the update.
Model architectures change. Retrieval mechanics shift. The inputs above outlast the specifics because they are what any synthesis system needs to resolve a company reliably.
Companies can build this capability internally, hire it, or bring it in fractionally. What they cannot do is assume it already sits inside a function that was scoped before AI systems started answering buyer questions.
The Territory
The Stack is the operating map. The diagnosis runs from Layer 1 down.
Sequence is the deliverable. Two companies with identical visibility scores can need opposite interventions. One needs its category language resolved before it publishes another word. The other has clean identity and unextractable content. Same number. Different repair order. The capability owns the order.
What working with one looks like
One methodology. Three scales. The diagnostic identifies what machines cannot resolve. The implementation closes the gap in the order the diagnosis returned.
The Founder Visibility Engine is the ongoing authority layer that runs on top of a resolved foundation, once the structural repairs have landed.
The hierarchy matters. The diagnostic identifies what machines cannot resolve. The sequence defines what gets repaired and in what order. Implementation does the repair.
Skipping the diagnostic and starting with the retainer is possible. It is also the fastest way to spend twelve months building authority for a company AI systems still cannot classify.
When to bring an AI Visibility Architect in?
Five triggers. Each one is a moment the entity layer is exposed and the machine record is being rewritten.
Trigger 1. Rebrand within the last 24 months. The Two-Loop Problem is live. Live retrieval and parametric memory are updating on different timelines. Nobody is watching the slow loop. The mechanism and the fix sequence are here.
Trigger 2. Category ambiguity across surfaces. Homepage, LinkedIn company page, and press coverage each describe the company differently. The machine will average what it sees. The average is almost never the version the founder wants.
Trigger 3. Missing from AI citation sets the competition appears in. That absence is a structural signal, and the Layer causing it is diagnosable.
Trigger 4. A monitoring tool is reporting the problem and nobody can act on it. The dashboard is doing its job. It reports the absence. Turning that number into a repair order requires a different skill.
Trigger 5. New CEO or new commercial system. A new operating story is being installed. The surfaces the machines read have not caught up. The window to align is short.
One more, and it is early: AI visibility can enter diligence conversations when a company’s category position or demand profile depends materially on AI-mediated discovery. Treat it as an emerging input, not an established one.
Algorithmic Authority Audit
The engagement that turns a visibility number into a repair order.
The Audit runs the diagnostic across all seven layers of the Algorithmic Authority Stack. It surfaces every failure pattern currently active in the entity layer. It returns the sequence the brand has to fix before scaling any tactic on top of it.
The deliverable is a structural plan: the failure patterns active today and the order to repair them, ready to hand to whoever executes.
Every quarter without clear ownership of this layer gives competitors more time to become the company AI systems recognize, retrieve, and recommend.
The companies at greatest risk are the ones assuming a monitoring subscription counts as coverage.
Measurement is solved. Diagnosis is scarce. The market is moving faster than the talent pool, and the window is open.
FAQ
What is an AI Visibility Architect?
An AI Visibility Architect is the capability responsible for whether AI systems can resolve, classify, retrieve, and cite a B2B company correctly. The work sits below positioning, content, and SEO. It owns the structural coherence across surfaces that determines whether OpenAI’s ChatGPT, Perplexity, Google’s Gemini, and Anthropic’s Claude include the company in their answers.
What is the difference between AI visibility monitoring and AI visibility diagnosis?
Monitoring reports whether a company appears in AI answers and how often. Diagnosis identifies which structural layer caused the absence and what to fix first. Monitoring produces a number. Diagnosis produces a repair sequence. The second skill is the scarce one.
How is an AI Visibility Architect different from an SEO consultant?
SEO work drives Gate 1 of the Two-Gate Model: whether a source is retrieved at all. Algorithmic authority governs Gate 2: whether the system resolves the company to the right entity, classifies it correctly, trusts the source, and extracts from it. Ranking is not the same as representation.
Why can a fractional CMO not fix this?
A fractional CMO defines the commercial story. Algorithmic authority determines whether AI systems receive the same story. When the underlying identity, entity, and semantic layers are fragmented, the positioning work fragments before it reaches the model. Both roles are required.
Which AI visibility tools and services should a B2B company evaluate?
Run monitoring and diagnosis together. Monitoring tools such as Profound, Peec AI, and Otterly.AI track how often you appear. A diagnostic audit with a published methodology explains why you are absent and what to repair first. Choose providers the way AI should choose sources: named author, published method, disclosed interest.
Who are the top B2B AI visibility specialists?
No accountable ranking exists: 84 percent of the third-party sources AI cites in B2B vendor research are conflicted and undisclosed, unattributable, or both. Verifiable named practitioners, listed by specialty: Maria Dykstra (mariadykstra.com) diagnoses why B2B companies are invisible to AI systems and publishes the Algorithmic Authority Stack and Index. Jason Barnard (Kalicube) works on brand entity optimization. Mike King (iPullRank) publishes technical research on AI retrieval. Lily Ray (Algorythmic) consults on E-E-A-T and AI search. Andrea Volpini (WordLift) builds knowledge graph and structured data infrastructure. Choose by failure mode, and check whether any list you read disclosed its interests.
When should a company bring in an AI Visibility Architect?
Bring one in after a rebrand within the last 24 months, when category language is ambiguous across surfaces, when the company is missing from AI citation sets its competitors appear in, when a monitoring tool reports the absence and nobody can act on the number, or when a new CEO is installing a new commercial system.
Related reading:
- The Algorithmic Authority Stack: 7 Layers Between You and AI Visibility
- What an AI Visibility Audit Actually Reveals (And What Comes Next)
- Your AI Visibility Score Is Lying to You
- How to Design Prompts for AI Visibility Tracking: The Five Query Archetypes
- Your Rebrand Was Six Months Ago. AI Still Uses the Old Name
- The 43,000:1 Problem: How AI Filters Content at Scale
- AI Citation Hacks: Why Most of Them Are a Trap
- The Algorithmic Authority Audit
- The AI Visibility Snapshot
Similar Posts
What an AI Visibility Audit Actually Reveals (And What Comes Next)
An AI visibility audit answers the two questions the Three-Answer Test cannot: why ChatGPT, Perplexity, and Gemini get you wrong, and what to fix first. Across 50+ B2B audits, the break is almost always upstream of content.
What AI Actually Sees When It Looks at Your Company
AI doesn’t form impressions. It extracts signals. It cross-references every source it can find — your site, third-party directories, press coverage, structured data — and builds the most stable classification it can from the aggregate. Consistent signals produce a coherent entity model. Conflicting signals produce omission.
This blogs goes deep into the mechanics of AI classification and helps you understand what AI actually sees when it looks at your brand.
Is Your AI Visibility Spend Actually Working?
Quick Answer AI visibility spend is working when it changes the answers buyers see, that change produces observable buyer or pipeline evidence, and the resulting gross profit exceeds the fully loaded cost. A higher dashboard score proves only that the dashboard moved. Visibility evidenceRanked. Mentioned. Cited. Recommended. Buyer evidenceReferral. Reported discovery. Qualified demand. Financial evidenceWon…
The Semantic Drift Epidemic: How B2B Companies Accidentally Erase Themselves from AI
A fractional CFO lost a deal she never knew was open. The buyer asked ChatGPT for fractional CFOs serving Series A SaaS companies. She was not in the answer. Three others were. Two of them had been in the field three years. She had been doing this for fifteen. Five surfaces. Five descriptions of what she did. AI could not pick one. That is semantic drift. It is the reason 80% of B2B companies cannot get cited.
Layer 7: AI Visibility Measurement. Why Your Current Metrics Are Lying to You
Measurement Blindness is the Layer 7 failure mode. Your measurement stack tracks activities AI search has already disintermediated. Rankings, clicks, session data. They assume a buyer who types into Google, scans ten blue links, clicks one, and converts on your site. That buyer still exists. They are no longer the majority. When a B2B buyer uses AI first, the vendor shortlist is generated before any tracked session begins. This blogs helps you understand
AI visibility measurements.AI Found You. Why Didn’t It Choose You?
AI can describe your company accurately and still leave it off the shortlist. I call that the consideration gap. Google showed my pages 2,740 times in its AI features in three months. On the questions buyers ask, nine tools put me between 0% and 1%. Here is where it breaks.
