AI gave your competitor a different answer on Thursday than it gave you on Tuesday.

AI gives different answers on different days. We don't. Every factor in the AI Ranking Factors framework returns the same true/false result, run after run — defensible, consistent, client-ready.

AI gives different answers on Tuesday than it does on Thursday. That's not a reporting system — it's a liability. Every factor in our AI Ranking Factors framework returns a deterministic true or false. Same result, every run, every week. That's what makes it defensible to a client.

AI assistants contradict themselves. Run the same prompt on Monday, get one answer. Run it Thursday, get another. That's not a measurement system — it's a coin flip with a confident voice. Every factor in our AI Ranking Factors framework is evaluated as a deterministic true/false test: the same answer, every run, week after week, auditable line by line. Out of 141 factors, not one is scored by asking an AI what it thinks. We use AI to surface opportunity. We don't let it grade your work. Your clients deserve a standard that holds — not one that drifts. — J. Brent Tuttle

Every factor we test returns a true or false answer. Not a score, not a sentiment, not a confident guess that changes by Thursday. True or false. That discipline exists because AI assistants hallucinate. They produce authoritative-sounding answers with built-in disclaimers about their own errors. They answer the same prompt differently depending on the day, the phrasing, or what they had for breakfast. That variability is a feature of the technology — and a catastrophic foundation for client work. The AI Ranking Factors framework tests 141 signals across the factors our analysis shows AI assistants actually use when deciding which local business to recommend. Each one resolves to a binary: either your business passes or it doesn't. No interpretation required. No "it depends." The same test run this week returns the same result as the test run six weeks ago, unless something about your business actually changed. That matters when you're sitting across from a client. You can show them exactly which factors they're passing and failing. You can fix a failing factor, run the test again, and show them it flipped. You can build a case that doesn't rest on a screenshot of a chatbot response that may never appear again. Most visibility frameworks in this space are built on top of AI outputs — essentially asking the machine to grade itself. We don't trust AI with your clients, and we don't think you should either. The recommendation engine is the thing we're trying to influence. It is not the thing we use to measure our own work.

Every factor we test returns a true or a false. Not a score. Not a confidence interval. Not an answer that shifts between Tuesday and Thursday. A binary result — present or absent — that is identical whether we run it today or six months from now. That discipline is not an accident. It is the entire point. AI assistants hallucinate. They produce fluent, authoritative-sounding answers that are factually wrong. They ship their own disclaimers about errors in the same breath they deliver those errors. Ask the same question on different days and you get meaningfully different responses. That is not a character flaw in the technology — it is a fundamental property of how large language models work. Probabilistic outputs are inherently variable. For a consumer asking where to get their car serviced, that variability is a mild inconvenience. For a business owner trying to understand why they are or are not being recommended, and what to do about it, that variability is worthless. This is why viizable applies a deterministic test to each of the 141 AI Ranking Factors. We are not asking an AI whether your business seems trustworthy. We are checking whether the structured data is present or absent. Whether the NAP record is consistent or inconsistent. Whether the technical signals that AI assistants draw on when they generate recommendations are actually there or not. Every test has one correct answer. Our analysis finds it, records it, and finds the same answer next week. The consequence is a client conversation that almost no one else in this space can have. When a factor shows false, you can point to exactly what is missing. When it flips to true after a correction, you can show the before and the after. You are not explaining a score that moved from 67 to 71. You are showing a client a list of conditions that were not met and now are. That is accountability. That is a defensible deliverable. Most visibility frameworks built around AI-driven search lean into AI tools to measure AI visibility. The problem is circular and compounding — you are using a probabilistic system to evaluate performance in a probabilistic environment, and the margin of error stacks. When a client asks why their score changed, you are explaining variance, not progress. The 141 factors in the AI Ranking Factors framework were identified through rigorous analysis of the signals that AI assistants consistently rely on when generating local business recommendations. The framework is built to be audited, not interpreted. Each factor is either satisfied or it is not. The audit result is reproducible. A competitor running the same check gets the same answer. There is no version of the truth that shifts with the model's mood. We do not trust AI to evaluate your client's visibility. We built something better — a structured, deterministic, repeatable test that gives you the same answer every time you need it. That consistency is not a limitation of the approach. It is the entire reason the approach holds up.

Every recommendation we make to a client is either true or false. Not probably true. Not true on average. Not true according to a model that processed your question on a Tuesday and might answer it differently on Thursday. True or false, full stop — and we can show our work. That discipline is not a technical quirk. It is the entire foundation of how we measure AI ranking factors for local businesses, and it is the reason we built our framework the way we did. Here is what prompted it. AI assistants — the systems that now answer consumer questions directly, recommend local businesses, and increasingly replace the first page of search results — hallucinate. That is not a criticism. It is a documented, structural property of large language models. They generate the most statistically plausible next token, not the verified next fact. They produce confident answers to questions they cannot actually answer. They ship their own disclaimers about potential errors in the same breath they cite those errors as authoritative. And they return different outputs for the same input depending on temperature settings, context windows, and the precise wording of the prompt. Run the same question on Monday, get one answer. Run it on Thursday, get another. Neither is necessarily wrong in any satisfying sense. Both are plausible. That is the point, and that is the problem. We are not criticizing the tools. We use them. We find them useful. But "useful" and "auditable" are different standards, and when you are making specific, consequential recommendations to a paying client about why their business is or is not being recommended by AI assistants — and what they should do about it — you cannot hand them a probabilistic output dressed up as an audit. So we did not. The AI Ranking Factors framework tests 141 discrete factors. Each one resolves to a binary outcome: pass or fail. Is the business name consistent across these specific data points? Pass or fail. Does the website load over a secure connection? Pass or fail. Is the primary category correctly mapped? Pass or fail. Does the structured data contain the fields that AI systems pull when constructing local recommendations? Pass or fail. There is no partial credit scoring that disappears when you run it again. There is no confidence interval on whether your NAP data is consistent. It either is or it is not. We check, we record, we return the same answer the next time we check because the underlying test is deterministic — the same logic, applied to the same observable fact, produces the same result. That reliability is not incidental. It is the thing. Consider what a client is actually buying when they engage with any visibility measurement service. They are buying an explanation for why they are or are not appearing in the answers that matter. They are buying prioritized guidance on what to fix. And they are buying accountability — the ability to compare results over time, to verify that the work done actually changed something, and to trust that the improvement they are seeing is real rather than an artifact of a model's mood on a given afternoon. You cannot deliver any of those things if your underlying measurement is variable. If the audit changes because the AI changed — not because the business changed — you have not measured anything. You have recorded a reading from a thermometer that runs warm on Thursdays. Every other major GEO and AEO visibility framework we have studied relies, at some point, on AI-generated assessments. They ask a language model to evaluate the quality of content, the relevance of a citation, or the apparent prominence of a business. Those assessments may be useful heuristics. They are not auditable facts. They cannot be reproduced on demand with the same result. They cannot be defended to a client in the specific, line-by-line way that client relationships eventually require. We have been in that meeting. The client asks why the recommendation changed from last month. The answer cannot be "the model saw it differently this time." That answer ends the relationship and it should. Our 141 factors were selected because they are testable, not because they are the most interesting or the most sophisticated. Some of them are unglamorous. Consistent phone number formatting across citations is not a thrilling topic. A well-formed schema markup block is not the kind of thing that generates enthusiasm in a strategy presentation. But these factors are the ones we can verify, the ones that AI systems demonstrably use when constructing local recommendations, and the ones where the pass-or-fail distinction actually holds. The framework produces a score. More important than the score, it produces a specific, ordered list of failures — the exact factors that are unresolved, ranked by their measured impact on AI recommendation probability. A client can take that list and work through it. They can check the work themselves. They can re-run the analysis and expect to see the same failures resolved and the same new state confirmed. The number moves because the facts moved, not because a model felt differently this week. We will be honest about what this does not do. It does not guarantee that fixing every factor will result in an AI assistant recommending your business. AI recommendation is probabilistic, contextual, and involves factors that no one outside the model developers can fully enumerate. We measure what is observable and testable. We improve what we can verify. We do not promise outcomes we cannot control. What we do promise is that when we tell a client something is true, it is true. When we tell them something changed, something actually changed. When we hand them a list of 141 factors and say six of them are failing, those six failures are real, reproducible, and addressable — not a snapshot of an AI's interpretation on the morning we happened to run the analysis. The entire value of measuring AI ranking factors rests on the reliability of the measurement. We are not willing to compromise that. We built the framework so we would not have to.

AI gives different answers on Tuesday than Thursday. That's not a framework. That's a coin flip. We don't use AI to evaluate your visibility. We run 141 deterministic true/false tests — same answer every single run, week after week. Your client deserves defensible data. Not confident guesses with a disclaimer buried at the bottom. Every factor in our AI Ranking Factors framework is measurable, repeatable, and something you can actually show a client without flinching. Scan. Don't guess. — J. Brent Tuttle #AIRankingFactors #LocalSEO #GEO

We don't trust AI with your client data. You shouldn't either. AI assistants hallucinate. They give confident answers with built-in disclaimers about errors — and answer the same question differently on Tuesday than on Thursday. That variability is not a foundation for client recommendations. Our AI Ranking Factors framework does something most visibility systems don't: it runs a deterministic true/false test on each of the 141 factors we track. Same answer every run. Week after week. Defensible in a client meeting. Measure what's real. Report what you can stand behind. Are the tools you're using built on answers that change — or on facts that hold? — J. Brent Tuttle

We don't trust AI with your client. You shouldn't either. AI assistants hallucinate. They answer the same prompt differently on Tuesday than on Thursday. They ship their own disclaimers about errors while sounding completely certain. That variability is not a basis for telling a client where they stand — or what to fix. The AI Ranking Factors framework tests 141 specific factors. Each one gets a deterministic true/false result. Not a score that drifts. Not a confidence interval. A clean, repeatable answer that looks the same on the next run as it did on this one. That's what makes it defensible to a client. Most GEO and AEO visibility systems lean on the same AI outputs they're supposed to be measuring. We test the underlying conditions that influence those outputs — structural signals, data consistency, authority patterns — using our own analysis and index. If you're going to tell a business owner their visibility is improving, you need a foundation that holds. Measure the factors. Not the hallucination. — J. Brent Tuttle

Why I stopped using AI to audit local business visibility (and what I use instead) I know this is going to be slightly controversial because everyone's excited about using LLMs to automate audits right now. But after running visibility work for local clients for a while, I had to make a hard decision: I don't trust AI to evaluate whether a business is visible or not, and I think a lot of people in this space are making a quiet mistake by doing so. Here's the core problem. AI assistants hallucinate. Not sometimes — regularly. They produce confident, well-formatted, citation-flavored answers that are factually wrong. They even ship with their own disclaimers telling you this. And the same prompt on Tuesday returns a different answer on Thursday. That variability isn't a quirk. It's a structural property of how these models work. Now apply that to client work. You run an audit. You tell a client "you're missing these signals" or "you're doing well on these factors." Next week, the model disagrees with itself. What did you actually measure? Nothing defensible. For anything that's going to inform a client recommendation — something they're going to act on, spend money on, or report upward — you need a deterministic result. True or false. Same answer every run. Not "the model feels like this is probably present." The specific things I'm talking about auditing are structured data, NAP consistency, schema implementation, page load behavior, entity clarity, citation accuracy — the kind of factors that actually influence whether an AI assistant recommends a business or surfaces it in a response. These are binary. Either the structured data is there or it isn't. Either the name matches across sources or it doesn't. Either the page responds cleanly or it doesn't. AI is genuinely useful for a lot of tasks in this workflow. Writing, summarizing, brainstorming content angles, drafting outreach. Fine. But using it as the measurement instrument itself? That's where I think people are going to get burned, and the clients are going to be the ones paying for it when the audit doesn't hold up three weeks later. Curious if others have hit this wall. How are you separating the "AI as tool" from "AI as judge" in your audits?

[COLD OPEN — tight on presenter, direct to camera, no music yet] Here's something most people in this industry won't say out loud: AI hallucinates. It gives different answers on Tuesday than on Thursday. It literally ships its own disclaimer that it might be wrong. So why are people using AI to grade your AI visibility? [PAUSE. Let that land.] If your agency is running a "GEO audit" and the audit itself is powered by a language model — you don't have an audit. You have a confident guess. And you cannot show a confident guess to a client and call it a strategy. [TITLE CARD: We Don't Trust AI With Your Client. You Shouldn't Either.] [SECTION 1 — THE PROBLEM] [B-ROLL: abstract visualizations of AI responses flickering, text changing, answers contradicting themselves on screen] Let's be specific about what variability actually means in practice. You ask an AI assistant which local plumber it recommends. Monday it says one name. Wednesday it says another. You ask why — it can't fully tell you. The weights shifted. The context window changed. Something upstream was updated. That's not a flaw. That's how large language models work. They are probabilistic by design. They are spectacular at generating language. They are not reliable scoring engines. Now imagine your GEO audit tool is built on that same foundation. You run it this week, you get a score of 74. You run it next week without changing anything — you get 68. Which one do you show the client? What do you tell them changed? This is the trap that almost every AI-visibility platform has walked straight into. And it's a trap that makes your agency look like it doesn't know what it's doing — even when the confusion isn't your fault. [SECTION 2 — THE DISCIPLINE] [B-ROLL: clean grid or checklist interface, binary indicators, true/false toggles resolving one by one] The AI Ranking Factors framework is built on one non-negotiable discipline: Every factor gets a deterministic true/false test. Not a score. Not a "likelihood." Not an AI-generated assessment of how well you're probably doing. True or false. Present or absent. Pass or fail. We test 141 factors across the signals that actually influence whether an AI assistant recommends a business. Things like: Is the structured data valid and correctly typed? Is the entity clearly defined and consistently represented across authoritative sources? Does the business appear in the data layers the models are trained on? Is the content formatted in ways that retrieval systems can actually parse? Each of those questions has a binary answer. Our analysis returns the same answer every run. That is not an accident — that is the point. [SECTION 3 — WHY THIS MATTERS TO YOUR CLIENT] [B-ROLL: presenter at a conceptual "client meeting" environment, light and professional] When you sit down with a client and show them their results, you need to be able to say two things with confidence: One — here is exactly what is working and what is not. Two — when we fix something, you will see it change in the next scan. That second one is where most visibility tools break down. If the baseline is fuzzy, the improvement is unmeasurable. If you can't prove the change, you can't justify the retainer. Deterministic testing means your baseline is locked. Week over week, the only thing that changes is what you actually changed. That's a defensible engagement. That's a report you can put in front of a CFO. [SECTION 4 — WHAT WE'RE NOT SAYING] [B-ROLL: presenter, calm and direct] To be clear — we are not saying AI assistants are bad. We use them constantly. They're extraordinary tools. What we're saying is: don't use a hammer to calibrate a scale. AI assistants are the environment your clients want to be found in. They are not the measurement instrument you use to figure out if your clients are found there. Those are two completely different jobs, and confusing them is costing agencies credibility they can't afford to lose. The AI Ranking Factors framework exists because the measurement job required a different kind of rigor. [OUTRO] [B-ROLL: presenter centered, calm close] If you're building an agency practice around AI visibility — or if you're a business owner trying to understand why your competitor shows up when yours doesn't — the foundation has to be honest measurement. Not AI grading AI. Not scores that drift. 141 factors. True or false. Same answer every time. That's the standard we hold ourselves to. And frankly, it's the standard your clients deserve. I'm J. Brent Tuttle. Subscribe if you want to keep following the AI Ranking Factors work — we're just getting started. [END CARD with subscribe prompt and link]

[Open on presenter, direct eye contact, no intro — speaking immediately] AI hallucinates. It gives confident answers with built-in disclaimers about those answers being wrong. It answers the same prompt differently on Tuesday than on Thursday. [Cut to slight push-in on presenter's face] That variability? That is not a foundation for a client recommendation. [Presenter steady, measured delivery] The AI Ranking Factors framework tests 141 factors — each one a hard true or false. Same answer every run. Week after week. [Pause — let it land] Deterministic. Defensible. Honest. [Closing — slight lean forward] We don't trust AI with your client's results. You shouldn't either. [End card / logo hold]

[COLD OPEN — presenter looks directly into camera, no intro, no warmup] AI hallucinates. It gives confident answers. Ships its own disclaimers. And asks the same question on Thursday that you asked on Tuesday — and gets a different answer. [slight pause, lean in] That is not a foundation for client recommendations. [cut or gesture shift] The AI Ranking Factors framework doesn't ask AI what it thinks. It runs a deterministic true/false test across 141 factors. Same result. Every run. Every week. [direct, slower close] Defensible to your client. Honest about what it measures. [beat] We don't trust AI with your clients. You shouldn't either.

[Presenter steps into frame, direct eye contact, calm authority] We don't trust AI with your client. [Beat — let it land] AI assistants hallucinate. They give confident wrong answers. They ship their own disclaimers about errors — and they answer the same question differently on Tuesday than on Thursday. [Slight lean forward] That variability is not a basis for a client recommendation. [Cut to clean graphic: "141 Factors. True or False. Every time."] The AI Ranking Factors framework runs a deterministic true/false test across 141 factors. Same answer every run. Week after week. Defensible. [Back to presenter] We're not scanning with AI. We're measuring what AI actually sees. [Pause, understated close] There's a difference.

[Cut to presenter, direct eye contact, confident — no intro, no smile yet] AI assistants hallucinate. They ship their own disclaimers about errors. They answer the same question differently on Tuesday than on Thursday. [slight lean forward] That variability is not a basis for client recommendations. [beat] Our AI Ranking Factors framework doesn't ask an AI whether your business is optimized. It runs a deterministic true/false test across 141 factors — same answer every run, every week. [direct, measured close] Defensible to your client. Built on what we can actually measure. That's not just a better method. It's the only honest one. [hold eye contact — fade]