How to run a Digital Reality Check
The Canvas maps the gap between what your company is, what it says, and what the internet can prove. The Check reveals it, using the same public evidence your buyers, search engines, and AI models see. You can run it in an afternoon, with no tools and no budget.
- What it is: a free, tool-free ritual to see the gap between what you are and what the internet can prove.
- How long: an afternoon, with your team in one room.
- The one move that matters: ask AI for the best options in your category, unbranded, and check whether you are in the answer.
- What you get: a one-page Canvas Readout with your gaps and next moves, not a score.
What the Check is (and what it is not)
A Digital Reality Check is a short, repeatable ritual. You ask AI and search the questions your buyers actually ask, you read your own public evidence the way a machine would, and you compare what you find to what you believe is true about your company.
It is a mirror, not a meter: it shows you the gap, but it does not put a number on it. The official, evidence-based measurement, the Digital Reality Score, is produced separately so the number holds up to scrutiny. The point of the Check is not to grade yourself. It is to feel the distance between your intended reality and your provable one, quickly enough that a whole team can see it in the same room.
Fill the Canvas (Start here)
The Check compares reality to something, and that something is the Digital Reality Canvas: a one-page template where you write down what you truly deliver, who you serve, what you claim publicly, and what evidence exists online. Spend thirty minutes filling it in with your team before you run a single query. The Check is only powerful because you have already committed, on paper, to what you think is true. Download the Canvas and fill it first.
Try one question first
Before the full walkthrough, ask a single unbranded question:
Ask AI: "What are the best options for [the problem your buyer is solving]?"
If your company is not in the answer, you have a gap. If you are in it but described wrong, you have a different one.
This one question is the whole framework in miniature. You are not asking the model what it thinks of you; you are asking it the question a real buyer asks, and watching whether the internet surfaces you at all. Most teams feel the gap immediately. The five steps below turn that flash of recognition into a full, shareable picture.
Use a fresh, logged-out session for every query — ChatGPT's Temporary Chat, an incognito window, or a signed-out browser — so memory and personalization don't hand you a flattering answer. You want the answer a first-time buyer gets, not the one your own history produces. Run each query from zero.
The full Check (5 steps)
Two of these steps test two different things that are easy to confuse. Keep them apart, because they fail for different reasons and need different fixes:
Machine Understanding
Whether AI can correctly classify and describe what you are.
"What does [company] do?"AI Representation
Whether, and how accurately, AI recommends you when buyers ask for options.
"Best options for [problem]?"Check Step 1: Ask AI (no brand)
Start where your buyers start, without naming yourself. Cover the whole buyer journey, not one query — you can be missing at some stages and present at others:
- Category research: what are the best options for [your buyer's problem]? Who leads [your category]?
- Comparisons: alternatives to [a competitor]; [competitor A] vs [competitor B].
- Use-case and regulation: what software helps with [the specific job, standard, or regulation you serve]?
- The thing you claim to own: ask directly about your core positioning claim — the capability or convergence you say only you do.
Write down the exact names that get cited, and in what order — not just "am I there." That competitor list is your baseline: re-run it next quarter and you can see who held the top spot, who slipped, and whether anyone new entered. This is your AI Representation and Recommendation: whether, and how, you are recommended when your name is not in the prompt.
Check Step 2: Ask AI (with brand)
Now put your name in. Ask what your company does, who it is best for, and what it is known for. This tests Machine Understanding: whether AI can correctly classify and describe what you are. A company can be understood perfectly and still never recommended, or recommended warmly but described with an outdated tagline.
Check Step 3: Search like a buyer
Run the same intent through a search engine. Search shows you which version of your company the web can retrieve and compare, and it often exposes the third-party pages shaping the AI answers above. Try:
- [company] + [category]
- [company] alternatives
- best [category] tools
- [buyer problem] stated plainly, as a buyer would type it
Check Step 4: Read your public evidence (website, reviews, 3rd party sites)
Open the pages a model reads about you, and read them coldly, as a stranger or a machine would, not as the person who wrote them:
- Your homepage and product pages
- Case studies and proof
- Reviews and directory profiles
- Marketplaces, partner pages, and comparison articles
For each one, ask: are the claims on my own site actually verifiable from the outside?
Check Step 5: Compare to find the Gap
This is the step that turns observation into insight. Put the answers you collected next to what you wrote in the Canvas. Every place the two columns disagree is a candidate gap, and it is now visible to everyone who ran the Check with you.
- 1. Ask AI (no brand) — are you in the answer? (AI Representation)
- 2. Ask AI (with brand) — does AI get you right? (Machine Understanding)
- 3. Search like a buyer — which version of you can the web retrieve?
- 4. Read your public evidence — is it verifiable from the outside?
- 5. Compare to find the Gap — where do the two columns disagree?
Why this works
The reason the Check lands is uncomfortable but useful:
- AI and search are rarely randomly wrong. They reflect the public evidence available to them.
- A miscategory usually has a source — an old profile, a thin homepage, or a competitor's comparison page the model read and trusted.
- You cannot prompt your way out of a gap. You close it by making the real you provable.
The uncomfortable truth is that AI increasingly recommends companies based on what the wider internet says about them, not only on what a company says about itself. The Check simply shows you the evidence the machines are working from, so you can fix the evidence rather than argue with the output.
See it in action
Here is the same Check run end to end on a fictional company, Northwind, which repositioned from a data-pipeline tool to a data-observability platform. When the team put the internet's answers next to their Canvas, the gap was specific:
That points to a Category & Positioning gap and an Outdated Reality gap, each with a clear fix. The full walkthrough shows every step, the AI and search results the team saw, the Canvas Readout, and the 90-day plan they built from it.
Read the full walkthrough, step by step →
Read out the gap
Close the session by writing a short Canvas Readout, not a score. Capture:
- Likely gap drivers — the root causes behind the mismatches you saw.
- Recommendation risks — where a buyer or a model is most likely to get you wrong.
- Your top three next moves — the changes to public evidence that would shift the answers first.
Keep it to a page. The value is in the shared language and the agreement, not the length.
You found the gap. Now what?
The Check is deliberately manual and directional: it shows you where the gap is, but it does not size it. The natural next step is to turn that read into a number you can act on and defend. When you want an evidence-based score, a benchmark against competitors, or ongoing monitoring as your evidence and the models keep changing, that is where PageRadar comes in, with the Digital Reality Score and Audit, and Digital Reality Radar to keep it live. The framework and the Check stay open and free; PageRadar is the measured layer on top.