AI VISIBILITY AUDIT
Measure how AI recommends your firm.
Buyers now put their shortlist questions to ChatGPT, Claude, Perplexity, Gemini, and Google's AI, and the answer comes back as a handful of names. The AI Visibility Audit measures how often yours appears, how you're described, and which competitors own the recommendation today.
October 3-7 | Starting at $500
Book a 30-minute walkthrough with Josh Miles
Founder of Bold Brand—brand strategy and marketing for AEC firms. Author of Bold Brand 2.0. Keynote speaker on technology, branding, and AI-era visibility for AEC audiences.
The AI search problem
These losses don't show up anywhere.
When buyers searched, you had evidence. Rankings, impressions, clicks: imperfect, but visible. You could tell when something was wrong.
AI answers produce no trail. A director of operations asks for a recommendation, gets four names, calls one. If you weren't in the answer, no system you own recorded that it happened. Firms are losing consideration weekly with zero signal that it's occurring.
And when the AI does talk about you, it speaks with total confidence about your services, your markets, your specialties, whether or not it's right.
Three things almost no firm can currently answer:
How often AI names you when a buyer asks without knowing your name
Which competitor is the default recommendation in your market
What the AI states about your firm that's false
The audit answers all three, and backs it up with data.
Why the test you already ran doesn’t count
Most principals have typed their firm's name into ChatGPT. That test is rigged in your favor three ways:
It's branded. Buyers who already know your name were never the risk. The unbranded question, "best geotech firm for transit work," is where new business lives, and it's the one you haven't tested.
You were logged in. Memory and chat history bend answers toward you. The result is flattering and worthless.
You ran it once. The same question, asked three times, returns three different lists. A single answer is an anecdote.
The audit tests the way buyers actually ask: unbranded, logged out, held to your market, repeated enough times that every result is a frequency.
What’s included:
Visibility score & tier. Invisible, Emerging, Established, or Dominant // scored from unbranded questions only, because unbranded is where new business comes from.
Share of voice. Your mentions divided by all mentions // a real percentage, benchmarked against every named competitor. This is the slide leadership remembers.
Competitive picture. Who owns the recommendation in your market, ranked, and normalized so you're only compared against firms of your own type.
Accuracy flags. Every false or outdated claim the AI makes about you, quoted verbatim, with the correction and the cost of leaving it standing.
Framing analysis. Not just whether you're named but how // the adjectives, hedges, and positioning the AI has assigned you, which is frequently not the positioning you built.
Source attribution. The specific rankings, directories, publications, and pages the engines cite in your category. This is the leverage map: change these sources and the answers change.
A ranked fix list. Four to six moves in priority order, each tied to the exact question and engine it should affect. Specific enough to hand to whoever executes.
A working session. We walk your team through the findings live and get you to a Monday-morning plan.
What it changes
Presence on the shortlist. AI answers name three or four firms. Moving from absent to present is the entire opportunity.
Misinformation, corrected at the source. An engine claiming a commercial-only firm "does residential" doesn't just misroute leads, it erodes the specialist positioning the firm spent years building.
A market map you can trust. You'll know which competitor the AI treats as the default, on which questions, and why.
A short list of high-leverage fixes. Source attribution usually reveals two or three publications doing most of the work in a category. That's a to-do list, not a year-long content program.
Rebrand verification. If your name changed, the engines are often still splitting your equity with the old one. Measurable, fixable.
A baseline while it's still early. Very few AEC firms are measuring this. The first firm in a market to manage it deliberately gets a head start that compounds.
Examples
Regional commercial contractor
Six buyer questions, five engines, three runs each. Named in 1 of 90 answers; two competitors held ~58% of recommendations. One engine invented a residential service line for a commercial-only firm. The fix list reduced to three source-level moves. (Anonymized real audit.)
National infrastructure engineering firm
Present in roughly a quarter of unbranded answers, nearly all of it from one engine. On the consumer engines their buyers use most, close to zero. The audit also caught an engine describing them as a division of a much larger competitor, a false claim handing brand equity to someone else, and found their pre-rebrand name still absorbing recognition two years after the change. (Anonymized real audit.)
Why Bold Brand?
20+ years of AEC brand strategy, the DRIVE process, the Bold Brand 2.0 book, leading from the SMPS stage. The buyer questions in your audit are written by someone who knows how architects, engineers, and contractors actually get hired, not generated from a keyword tool.
Custom-built for AEC firms
A great fit for: AEC firms competing in a defined market or specialty · firms post-rebrand or post-merger checking what carried over · marketing leads who need a defensible number to put AI search on leadership's agenda · firms watching a competitor surface in conversations they used to own.
Not for: traditional keyword-and-backlink SEO audits · execution of the fixes (a separate engagement we can scope) · anyone who wants a guaranteed position in AI answers… that guarantee doesn't exist, and nobody selling it can deliver it.
What we won’t claim
The audit is a snapshot. AI answers shift as sources get indexed and models update. One audit tells you where you stand and what to fix; the trendline comes from re-running it. That's what quarterly monitoring is for: a single data point isn't a trend.
Untested is not absent. When an engine can't be tested cleanly, we mark it untested and disclose it. We don't estimate, and we don't blur the two to improve a chart.
Every number is a frequency. "Named in 4 of 15 answers," never "ChatGPT recommends you." One run is never a finding.
FAQ
How long does it take? About three weeks from brief to working session.
What do we need to provide? A 45-minute intake, your actual services and service area (accuracy scoring depends on the truth), and 3–10 competitors, or we infer them.
Which engines do you test? ChatGPT, Perplexity, Gemini, Claude, and Google's AI Overviews.
Can you guarantee we'll show up? No, and no one honestly can. We can tell you precisely why you don't today and which changes are most likely to move it.
How is this different from SEO? SEO competes for placement on a page of links. This measures presence inside the answer itself, the discipline is Generative Engine Optimization. The two connect: your audit's source attribution tells any SEO effort exactly which sources matter.
We already have an SEO agency. Keep them. Hand them the report. It's a target list.
How often should we re-run it? Quarterly. Long enough for changes to register, short enough to catch a competitor's surge or a new hallucination early.
Thirty minutes, one live answer.
On the call, we ask an AI a real question from your market and read the result together. If you're named, you'll know and you'll have gotten a free data point. If you're not, you'll have watched the problem happen in real time, and you'll know exactly what the audit measures.