Your AI Visibility Consultant Should Be Visible to AI
I ran a simple test. I opened a consultant's site and my own in a plain text browser, the kind that strips away every photo, font, and color and shows only what a machine actually reads. Same tool, same settings, both sites.
Their homepage rendered to two screens of text. Mine rendered to eight (re-checked today; same numbers as the first time). Roughly four times the readable surface, on a site with a MOZ DA score of 7 out of 100. That is the domain authority score, a third-party SEO measurement that estimates how well a site might rank, scaled 0 to 100. Mine is in the single digits.
That gap is the whole story. The other site was not broken. It read clean. It had a sharp value proposition and a tidy list of services. What it did not have was anything for a machine to stand on: no named clients, no case studies with real outcomes, no methodology a model could point to and cite. It was legible, and it was empty.
Mine was not bigger because I wrote more for the sake of it. It was bigger because every claim has a name attached. Real clients. Real results. Real pages a model can read, verify, and quote. That is the difference between being listed and being recommended, and you can see it in plain text before you see it anywhere else.
Key Takeaway
A consultant's site can read perfectly to a machine and still give it nothing worth repeating. Legible is not the same as recommendable. The real test is not whether your name appears, it is whether the page hands an AI model something to stand on: real names, real results, real proof it can quote with confidence.
Roughly four times the readable surface. On a site with a fraction of the domain authority.
A second piece of evidence from the same week: my redpingeek.com page showed up on the same Google search results page (the SERP, in SEO terms) as a Mastercard page for an agentic-commerce search in my category. Mastercard's site carries a MOZ DA of 91 out of 100. Mine carries a 7. Same search, same first page of results.
A third, and the strongest of the three: I asked Perplexity (one of the AI search tools that pulls from web pages to build answers, the same way ChatGPT can) to tell me about agentic commerce for the jewelry industry. It gave me an answer. Then it showed me the list of websites it said it used to build that answer. My redpingeek.com page was on the list. So were Stripe, Mastercard, Salesforce, McKinsey, ACI Worldwide, Juniper Research, Deloitte, and PwC, all enterprise publishers carrying domain authority scores in the nineties on Moz's 0-to-100 scale. When I read the answer Perplexity gave, it was quoting from my own published content throughout the response. The AI was not just listing my site. It was using my pages to build what it was telling people about my industry.
I had first asked this same question seven weeks earlier, on April 14, 2026, and Perplexity had cited my pages then too. I re-ran it again on June 3 and June 4 to verify. Three captures across nearly two months. The order of the websites in the source list moved around between them. But mine was still there each time. And the AI was still pulling from my pages to build its answer each time. The point is not whether I was first or fifth on the list. The point is that the AI keeps coming back to my pages for the substance of what it says, across at least one Perplexity model cycle, even when those huge-brand sites have MOZ DA scores in the nineties and mine sits at 7.
MOZ Domain Authority scores on a 0-to-100 scale. Same answer. Same source list. Very different domain authority.
Domain authority does not predict whether AI or Google surfaces you for a specific question. The technical foundation underneath the page does. Three independent checks (the plain-text render, the Google SERP, and the Perplexity source list), all pointing at the same conclusion.
What "legible but thin" actually looks like
I have been running a version of this check on consultant sites for months. Sometimes a client asks me to look at a consultant they are considering. Sometimes I find one by accident. The pattern is consistent enough now that I want to name it.
The legible-but-thin site reads cleanly. Headings are in place. Service modules are listed. There is usually a stat block somewhere making a confident claim about how many businesses they have helped or how much revenue they have generated. The visual design is professional. None of that is a problem.
The problem is what happens when you read the page the way a model reads it. The stat is unattributed. The service list is generic enough to belong to any consultant in the category. There are no named clients you could verify. There are no case studies a model could quote. The "trusted by industry leaders" line points to logos that may or may not be real, and there is no narrative anywhere on the page about what working with this person actually involved.
This is the invisible middle made concrete. The page is not broken. It is published. It is technically indexable. A model can crawl it and read it. What the model cannot do is recommend it with any confidence, because there is no concrete claim on the page that the model can point to. Every sentence is a promise. Nothing is evidence.
Dr. Marie Haynes, an SEO and AI search authority whose work I follow closely, put it plainly:
Google's goal is not to make websites rank.
Dr. Marie Haynes, SEO and AI search authority
Her point is the one this piece keeps circling: the system serves the answer, not the site. An Avidly Agency analysis of the new search landscape backs it up: only 7.2% of domains show up in both Google's AI Overviews and the source lists of AI tools like ChatGPT and Perplexity. The two worlds reward different things, and the one that wins AI citations rewards depth and proof, not domain age.
The other 92.8% are split. One set of domains wins Google. A different set wins AI. Different incentives, different reward systems.
By contrast, when I run the same check on a site that does the work, the structural difference is immediate. Named clients appear in the navigation. Each named client links to a case study with a date, a context, and an outcome. The methodology is described somewhere on the site in enough detail that a model could quote a sentence and stand behind it. The first line of each page states plainly what the page is, not what the brand wishes the page felt like.
You do not need a plain text browser to see this. You just have to know what to look for.
The five-step check you can run without any developer tools
This is the test. Five steps. Each one uses something you already have: a browser, an AI tool, your own attention. You can run all five on any consultant's site in about ten minutes.
Step 1: The AI read-back test
Paste the consultant's homepage URL into ChatGPT, Claude, or Perplexity, and ask: Read this page and tell me what this business sells, who is behind it, and who it is for.
If the answer is vague, hedged, or wrong, the page is not communicating clearly to a machine. This is not a clever trick. It is the closest thing you have to a stand-in for what a crawler sees. The AI is reading the same words, the same structure, the same emphasis. If the AI cannot summarize the page back to you with specificity, the page is not giving a model anything specific to repeat.
Pay attention to what the AI does NOT say. If it cannot name the consultant, that is a signal. If it describes the offering in generic category language ("an SEO consultancy that helps businesses get found online"), the page did not give the model anything more specific than the category itself. If it makes the consultant sound interchangeable with any other consultant, the page made them interchangeable.
Step 2: The Reader Mode test
Open the page in a browser and switch on the built-in reading view. Safari: tap the "aA" icon in the address bar and select Show Reader. Chrome and Edge: there is a reading-mode option in the menu (sometimes you have to enable it once in settings, then it appears as an icon).
Reader Mode strips away the design and shows the page as plain text. This is the closest non-developer stand-in for the plain text browser I used in my own diagnostic. If the page collapses to almost nothing in Reader Mode, the content was carried by the design, not by the text. If key claims vanish, those claims were rendered as images or as styled flourishes the reading view does not surface.
The consultant whose page I rendered down to two screens of text? Their page is dense and active in a browser. The animation alone is doing real work to make the page feel substantial. In Reader Mode, none of that holds. What remains is what the model would see. A model does not see your animation.
Step 3: The names-and-proof scan
Read the page with one question in mind: where is the concrete proof?
A consultant who does the work will have:
- Their own name and credentials visible, not hidden behind a generic brand voice
- Named clients you can independently verify
- Case studies with dates and specific outcomes, not just adjectives
- A methodology described in enough detail that you could rephrase it
Hollow authority looks like the opposite of all of those:
- Generic brand voice with no human attached
- "Trusted by industry leaders" with logos that may or may not represent real engagements
- Results stated as percentages with no source ("3x your traffic" without saying whose, when, or how)
- Claims of expertise without any narrative of how the work actually happens
This is the difference between being listed and being recommended. Listed means a machine can find your page and read it. Recommended means a machine, asked who to suggest in a category, has enough to stand behind the suggestion. The bridge between those two states is the named, dated, attributed proof on the page. If it is absent, you are looking at the invisible middle.
Step 4: The "say what it is" test
Read the first line of each important page on the consultant's site. The homepage. The about page. The services page. A blog post.
Does the first line state plainly what the page is and who it is for? Or does it open with mood and stats and brand atmosphere?
Crawlers and models decide page relevance from the top down. They weight the first lines more heavily than the rest. A page that opens with "Empowering businesses in the AI age to unlock their full potential" tells a model very little, because every consultant on the internet could open with that line. A page that opens with "I help independent jewelry designers get cited by AI tools like ChatGPT and Perplexity" tells a model exactly what the page is.
Pay attention to whether the consultant is writing for the person who will read the page or for the search engine they think they need to impress. The first-line test surfaces that distinction in about ten seconds.
Step 5: The free-tool check
When you have run the first four steps and have a feeling about the consultant's site, point your own store at the free tools that do the same audit at a structural level: AI Page Visibility Check and the AI Visibility Score. Both are free. Neither requires installing anything. They tell you what your own store looks like to AI tools, which is the version of the diagnostic that actually matters for your business.
If the consultant's site passed your first four steps and your store needs work, you have a candidate. If the consultant's site failed your first four steps and your store needs work, the consultant cannot help you. They have not been able to help themselves.
When the basic check tells you to look closer
The five-step check above is the version any designer can run in a browser, with no installation and no developer tools. It is enough to identify the legible-but-thin pattern in most cases, and most of the time the answer it gives you is the right one.
There is a second layer of signals that I check when the basic version gives me a borderline answer. Whether the consultant has a live llms.txt file (a small text file that tells AI tools how to read the site, the way robots.txt used to tell search engines how to crawl it). Whether their service pages have valid schema markup (the behind-the-scenes structured data labels that tell search engines and AI tools what a page is about; you can check this with Google's free Rich Results Test). Whether their sitemap.xml, the map of pages a site declares to crawlers, actually loads with a real list of URLs instead of returning a 404 (page-not-found error). Whether they get recommended by name when I ask ChatGPT to suggest consultants in their stated specialty.
Those signals are not visible to the naked eye and they are not visible in Reader Mode either. They are the technical foundation underneath the readable surface. A consultant who has them in place is investing in the same architecture they should be building for clients. A consultant who is missing all of them is asking you to trust them with infrastructure they have not been able to build for themselves.
I put the deeper check inside the Consultant Evaluation Scorecard because the technical signals matter and most designers cannot easily check them on their own. The tool walks you through the deeper checks one click or one paste at a time, with plain language for what each result means. You never see a validator output. You never run a terminal. You see "did this URL load when you clicked it" and the tool does the translation.
That is the Premium layer. The blog post is the free version. Both ask the same question, just at different depths.
Listed is not recommended
The pattern that started this piece is the pattern I run into every time I get a referral from a designer who has been burned. A consultant came highly recommended. Their LinkedIn looked impressive. Their domain authority was high (in the conventional SEO sense; meaning their site had been around a long time and built up a backlink profile). Their site loaded fast and looked professional. The designer engaged them. The work happened. The result was nothing.
When I went back and ran the trust signal check on the consultant's own site, the answer was always the same. The site read clean. It was legible. It was also empty. There was no named client to verify. No case study to quote. No methodology to point to. No record of the consultant being recommended for their own stated specialty.
The lesson is the one this piece tries to surface: the trust signals you should be checking on the consultant's site are the same signals they should be building on yours. If they cannot build them on their own, they cannot build them on yours either. The diagnostic is symmetrical. Run it on them before you let them run it on you.
If you want a second opinion on a specific consultant or your own store before any engagement, the AI Visibility and Agentic Commerce Audit is the private read. I run the same check I ran in this piece, against the same standard, and tell you what I find. Snapshot tier starts at $97; the full audit and the full audit with a 1:1 call sit above that.
The free check is yours either way. Use it.
Frequently Asked Questions
Sources
Continue Learning
- Fashion, Demi-Fine, or Fine: How Each Buyer Researches: the buyer-tier framework. If the trust signal paradox is about evaluating consultants, the buyer-tier piece is about evaluating which buyer your own store is signaling to.
- New Platform, Same Problem: Why Switching to Shopify Won't Fix Your AI Visibility: the same "visible fix" pattern applied to platform migration. Content architecture, not the visible swap, is what determines visibility.
- Score your own store on the free AI Visibility Score: the diagnostic that turns "listed vs. recommended" into a measurable number. Same trust signals from this piece, applied to your site. Takes about 30 seconds.
- Subscribe to the Red Pin Geek Substack (free): weekly inside-view content for jewelry designers building AI visibility on their own. Premium tier unlocks the Consultant Evaluation Scorecard, the Content-to-Buyer Tier Audit Worksheet, and the rest of the Premium tool series.
- AI Visibility and Agentic Commerce Audit (Snapshot $97 / Full Audit $597 / Full Audit + 1:1 call $997): the private read. I run the same five-step check on your store and tell you what I find.
About Andrea
Andrea Li is the founder of Red Pin Geek, an AI visibility consultancy for independent jewelry designers. She has spent eighteen years in the jewelry industry, including running her own jewelry brand at andreali.com, where she has documented the convergence between traditional SEO and AI citation eligibility. Red Pin Geek is the practitioner-led methodology she developed first on her own store, then refined through client engagements including Bohemi and Talisman Fine Jewelry. Andrea writes for the jewelry designer at the bench, not for the SEO professional at the desk.
Want Red Pin Geek in Your AI Answers?
Google now lets you choose preferred sources that get labeled in AI Overviews and AI Mode. If this content helps you navigate AI visibility for your jewelry business, add us and you'll spot our links faster next time Google answers your questions.
Add Red Pin Geek as a Preferred SourceTakes one click. Works in Google Search, AI Overviews, and AI Mode.
