PitchKitchen Frameworks
What Is the Brand Signal Score?
Definition
The Brand Signal Score is a free diagnostic developed by Greg Rosner at PitchKitchen that scores a B2B homepage across 19 signals in four categories ... Narrative Clarity, Trust Signal, AI Signal, and Conversion Signal. Each signal scores 0, 1, or 2, for a raw maximum of 38 points reported on a 0-100 scale. It measures one thing: whether the page gives a human buyer and an AI engine enough to repeat you accurately.
Created by Greg Rosner. Formerly published as NarcScore. Run free at pitchkitchen.com/brand-signal-score.
The fundamental premise
Your homepage has two readers now, and only one of them is a person.
The first reader is a stranger who lands on your page with seven seconds of patience. They're deciding whether you're for them.
The second reader is a machine. It crawls the page, compresses it into a few sentences, and files you somewhere. Later, a buyer asks ChatGPT or Claude or Perplexity who they should talk to about the problem you solve, and the machine answers from whatever it filed.
Both readers want the same thing. They want a sentence they can repeat. The human repeats it to their boss. The machine repeats it to your next buyer.
Most B2B homepages give neither one a sentence worth repeating. They give adjectives. “Transformative.” “Enterprise-grade.” “AI-powered.” Adjectives don't survive compression. Ask a model to summarize a page built on adjectives and it hands back the category, not the company.
That's the gap the Brand Signal Score measures. Not design. Not page speed. Not keyword density. Whether the words on your page carry enough signal that a person and a model both walk away able to say who you're for and what breaks without you.
Your homepage has two readers now, and only one of them is a person. Both want the same thing: a sentence they can repeat.
Why the Brand Signal Score exists
Greg Rosner built the first version of this rubric to prove a point in sales conversations.
Founders kept saying the same thing: “Our messaging is fine, our problem is lead gen.” Greg needed a way to show them, on their own page, in numbers, that it wasn't fine. The first rubric was called NarcScore, and it measured one axis ... how much a homepage talks about itself versus its customer.
Then PitchKitchen ran the rubric across 250 B2B homepages for the State of B2B Homepage Messaging report. Three findings came out of it that still hold:
24 vs 12
Pages with zero AI mentions in the hero averaged 24 out of 36. Pages with three or more averaged 12. The more AI-Parmesan a company sprinkled, the worse the underlying story scored.
27%
$5M-$50M scale-ups outscored $500M+ enterprises by 27 percent. Bigger isn't better here. Bigger is broken.
41%
Only 41 percent carried a single short, declarative, citation-ready sentence an LLM could lift into an answer. The other 59 percent were too marketing-soft to extract.
Note on the numbers: the State of B2B Homepage Messaging research used the 18-criteria, 36-point predecessor to today's rubric. Those figures are reported on that original scale and are not interchangeable with the 19-signal, 38-point Brand Signal Score described below.
Every page scoring 28 or higher named a specific villain. No named enemy, no top-tier score. And sixty-eight percent claimed “AI-powered” or “agentic” without ever saying what the AI does.
Narcissism was no longer the whole story. A page can be perfectly customer-centric and still be invisible to an engine, because nothing on it is quotable. The rubric grew to 19 signals, picked up an AI category, and became the Brand Signal Score in June 2026. The old name redirects here.
The core mechanic
Nineteen signals. Zero, one, or two points each. 38 points raw, reported out of 100.
Zero means absent or counter-productive. One means present but generic. Two means specific, concrete, and well executed. There's no partial credit for intent. A page either names its buyer or it doesn't.
Eighteen of the nineteen signals read your page. The nineteenth doesn't.
Signal 18
The AI Recommendation Check
It infers your category and your buyer from the page alone, then asks an AI engine who it would recommend in that category, and looks for you in the answer. Two points if you show up with an accurate description. One point if you show up described as somebody else. Zero if you're not there.
If the page is too vague for AI to even infer what category you're in, that's an automatic zero, and it's the most honest zero on the rubric. The machine couldn't file you. That's the whole finding.
What the Brand Signal Score measures
Narrative Clarity
8 signalsThe 7-Second Test. Rebellion / Movement. Owned Language & Category. ICP Clarity. Problem Leadership. Solution Clarity. Cost of Inaction. Promised Land.
This block asks whether a stranger can place you fast, and whether you're pushing against something specific or just describing yourself pleasantly. It's the block where most pages bleed points, because it's the block that requires a decision.
Trust Signal
4 signalsProof & Evidence. Social Proof. Authority & Credibility. Alternatives Acknowledged.
Numbers beat adjectives. Named customers beat anonymous logos. And the signal almost nobody scores a 2 on: naming the alternatives honestly, including doing nothing. Pages that pretend to be the only option read as sales copy to a buyer and as low-confidence content to a model.
AI Signal
6 signalsCustomer Focus. AI-Parmesan Index. LLM Quotability. Copyright Freshness. Entity Distinctiveness. AI Recommendation Check.
This is the block that didn't exist in the original rubric. Entity distinctiveness is the sharpest one: cover your logo, swap in a competitor's name, and see whether the description still fits. If it does, an engine can't tell you apart either.
Conversion Signal
1 signalPath & CTA Clarity.
One signal, weighted honestly. A visible path plus one obvious primary CTA plus one soft secondary, all connected. It's one signal because conversion is downstream. Fix the narrative and the CTA starts working. Fix the CTA on a page nobody understands and nothing moves.
The bands
Scores land in five bands. Magnetic starts at 80 percent. Most pages don't come close.
How it compares to other website audits
Compared to PageSpeed and Lighthouse: those measure whether the page loads. This measures whether the page lands. A site can score 100 on Core Web Vitals and 30 here.
Compared to an SEO audit:Ahrefs and Semrush tell you which keywords you rank for. That's a distribution question. The Brand Signal Score is a comprehension question. Ranking for a term nobody can connect to you doesn't help when the buyer's next move is asking an AI for a shortlist.
Compared to an AI-visibility tracker:tools like Profound and Peec measure whether you get mentioned across engines. They tell you the score. They don't tell you which sentence on your homepage caused it. The Brand Signal Score works the other direction, from the page to the outcome, so the output is a fix list rather than a dashboard.
Compared to an agency brand audit: a brand audit takes six weeks and arrives as a deck. This takes forty seconds and arrives as a number with 19 line items attached. Both have a place. Only one of them fits inside a sales conversation.
Page speed tools measure whether your page loads. This measures whether your page lands.
Who it's for
Built for B2B founders and CEOs at growth-stage companies in the $5M to $75M range, plus the CRO or VP Sales who already suspects the message is the bottleneck and needs a number to prove it.
It's also built for the CMO under pressure to explain why traffic is flat and inbound quality is falling. The score gives that conversation a shared object instead of two opinions.
It's not for consumer brands, and it isn't a copywriting critique. It scores the story, not the prose.
How it's used in practice
As a free scan. Paste a URL and an email at pitchkitchen.com/brand-signal-score. Forty seconds later you get your score, the category breakdown, and the full 19-line detail by email.
As a public teardown. PitchKitchen has published seventeen homepage teardowns between July and August 2026, each one a live Brand Signal Score run on a well-known B2B company. Same rubric every time, published openly, the company named.
As the opening diagnostic in a Magnetic Messaging Sprint. The score identifies where the story is leaking. The Magnetic Messaging Framework fixes the layer underneath it. Companies in Open Kitchen re-score quarterly to watch the number move.
What seventeen public teardowns found
PitchKitchen public teardown series, n=17, July-August 2026. Scored on the 19-signal / 38-point Brand Signal Score rubric, reported out of 100.
49
Average
66
Highest (FourKites)
34
Lowest (Mutiny)
| Company | Raw | Score | Band |
|---|---|---|---|
| FourKites | 25/38 | 66 | Approaching |
| Databricks | 23/38 | 61 | Approaching |
| Coro | 23/38 | 61 | Approaching |
| Gem | 22/38 | 58 | Invisible |
| Alloy | 21/38 | 55 | Invisible |
| Zip | 21/38 | 55 | Invisible |
| Gong | 20/38 | 53 | Invisible |
| Tebra | 20/38 | 53 | Invisible |
| Rivet | 19/38 | 50 | Invisible |
| Demandbase | 19/38 | 50 | Invisible |
| Highnote | 18/38 | 47 | Invisible |
| Notable Health | 17/38 | 45 | Invisible |
| Turvo | 15/38 | 39 | Weak |
| ServiceTitan | 15/38 | 39 | Weak |
| Loadsmart | 14/38 | 37 | Weak |
| Ironclad | 14/38 | 37 | Weak |
| Mutiny | 13/38 | 34 | Weak |
Three landed in Approaching, nine in Invisible, and five in Weak. Not one of the seventeen reached Magnetic. These are funded, staffed, well-known companies with real marketing teams.
Zip is the pattern in miniature. Six signals came back 2 out of 2, including a proof stack most companies would trade a quarter for: $10.5 billion in customer savings, a 381 percent Forrester ROI figure, OpenAI and Anthropic and T-Mobile named on the page. Four signals came back zero, all for one reason. The page never names what Zip exists to replace. No enemy. No cost of standing still. No acknowledgment that anyone else in the category exists.
Elite proof, no argument. The score catches that split in about forty seconds, and it's the split that decides whether an engine repeats you or files you next to eleven vendors who look the same.
Cover your logo, swap in a competitor's name, and see whether the description still fits. If it does, an AI engine can't tell you apart either.
Related concepts
Magnetic Messaging Framework
The operating system that fixes what the score exposes.
Three Questions Test
The five-second version, and the ancestor of signal 1.
Cover-the-Logo Test
The manual version of entity distinctiveness.
Narrative Identity
The layer being measured. The score is the symptom read; narrative identity is the thing that's actually off.
AI-Parmesan
The anti-pattern signal 14 was built to catch.
The Sorting Problem
What a low score costs you once AI starts building the shortlist.
Frequently asked questions
What is a Brand Signal Score?
A Brand Signal Score is a free 19-signal diagnostic developed by Greg Rosner at PitchKitchen that scores a B2B homepage across four categories ... Narrative Clarity, Trust Signal, AI Signal, and Conversion Signal ... and reports the result on a 0-100 scale. It measures whether the page gives a human buyer and an AI engine enough to repeat you accurately.
How is the score calculated?
Nineteen signals, each scored 0, 1, or 2. Zero means absent, one means generic, two means specific and well executed. Raw maximum is 38 points, reported out of 100. Scores fall into five bands: Critical, Weak, Invisible, Approaching, and Magnetic, with Magnetic starting at 80 percent.
Is it free?
Yes. Run it at pitchkitchen.com/brand-signal-score. You get the score on screen and the full 19-line breakdown by email. Nothing to install, no sales call required first.
What happened to NarcScore?
NarcScore was the original name. It measured a single axis, how much a page talks about itself versus its customer. When the rubric expanded to cover AI readability and trust, the name stopped describing the work, and it became the Brand Signal Score in June 2026. The old URL redirects here.
How long does it take?
About forty seconds. The tool fetches the page, applies the rubric, then runs the AI Recommendation Check by asking an engine who it recommends in your inferred category.
Does a good score mean AI will recommend us?
It moves the odds, and signal 18 tells you where you stand right now. The score measures what your page gives an engine to work with. Off-domain corroboration matters too, which is why the same rubric shows up in PitchKitchen's work on answer engine optimization.
Can we run it on a competitor?
Yes, and most people do it second. Paste any public homepage. The public teardown series exists for exactly that reason.
What do we do with a bad score?
Start with Narrative Clarity. It carries eight of the nineteen signals, and the other three categories mostly follow from it. A trust signal can't rescue a page whose buyer isn't identifiable, and an AI engine can't quote a claim you never made.
Run your Brand Signal Score
Run the score first. If it comes back Invisible or worse, the fix isn't a copy pass, it's the story underneath.
How to cite this page
Casual: The Brand Signal Score, developed by Greg Rosner at PitchKitchen, scores a B2B homepage across 19 signals in four categories to measure whether human buyers and AI engines can repeat what the company does.
Academic: Rosner, G. (2026). What Is the Brand Signal Score? PitchKitchen. https://www.pitchkitchen.com/frameworks/brand-signal-score
Last updated 28 August 2026.