The checker scores a page on the things search engines and AI assistants actually read. This page explains every one of them: what it is, why it matters, and how to fix it. Where a figure appears, it comes from our random-sample study of 2,500 Richmond-area businesses — so you can see how common each problem really is locally.
Being readable by machines
Structured data (Schema.org)
Structured data is a block of machine-readable facts about your business — name, address, phone, hours, services — embedded in the page as JSON-LD. Humans never see it; search engines and AI assistants read it first.
Without it, a machine has to infer who you are from your prose, and it often infers wrong or gives up. With it, there is no ambiguity about which business the page is describing.
The fix: add a
LocalBusiness or Organization block to every
page, and a Service block on each service page. The details
must match your Google Business Profile exactly — a mismatch is worse
than nothing.
llms.txt
A plain text file at /llms.txt that tells AI
crawlers, in their own preferred format, what your site is and which
pages matter. It is the newest convention here and the least adopted.
It is not a ranking factor and nobody should claim it is. What it does is remove guesswork: instead of an assistant assembling an impression of your business from whatever it happened to crawl, you hand it a summary you wrote.
The fix: a short Markdown file with your business in one sentence, what you do, the pages worth reading, and how to contact you. Ours is at tools.rvadigitalworks.com/llms.txt if you want a template.
AI crawler access
ChatGPT, Claude, Perplexity and Google's AI features each
crawl with their own named bot — GPTBot, ClaudeBot, PerplexityBot,
Google-Extended. Your robots.txt decides whether they are
allowed in.
Plenty of sites block them by accident, usually because a plugin or a host added a blanket rule. If a crawler cannot read your site, no amount of good content will get you cited by the assistant behind it.
The fix: name the AI crawlers explicitly
in robots.txt and allow them. Blocking is a legitimate
choice — but it should be a choice, not an accident.
Being quotable
Answer-first content
Whether your page answers the question it is about in the first sentence or two, rather than building up to it.
AI assistants quote passages, not pages. A paragraph that states a fact plainly can be lifted into an answer intact. A paragraph that opens with "In today's competitive landscape..." cannot, because there is no sentence in it worth quoting.
The fix: put the answer first, then the context. If someone asks "how much does a kitchen remodel cost in Richmond," the page should say what it costs before it says anything else.
Title and meta description
The title is the strongest single signal about what a page is for. The meta description does not affect ranking, but it is often what a person reads before deciding to click.
The common failure is a title that names the business but not the service or the place — "Home | Smith Co" tells a machine nothing about what Smith Co does or where.
The fix: service, then location, then business name. "Kitchen Remodeling in Midlothian, VA | Smith Co" works for both a person and a machine.
Heading structure
One h1 per page describing the page, with
h2 and h3 marking real sections beneath it.
Headings are how a machine segments your page into passages. When headings are used for visual size rather than structure, the page reads as one undifferentiated block and nothing in it can be quoted cleanly.
The fix: use headings for hierarchy and CSS for size. Phrase them as the questions people actually ask.
Being usable
Mobile rendering
Whether the page declares a viewport, which is what tells a phone to render it at phone width instead of shrinking a desktop layout down.
Most local searches happen on a phone. A site without this is not merely ugly on mobile — it is close to unusable, and Google has indexed the mobile version of pages first for years.
The fix: one line of HTML in the
head. If it is missing, that usually signals the site
predates responsive design and needs more than the one line.
HTTPS
Whether the site serves over a valid certificate rather than plain HTTP or an expired cert.
Browsers show a "Not secure" warning next to the address, which costs trust at exactly the wrong moment. Expired certificates are the more common cause locally — someone set it up once and the renewal lapsed.
The fix: certificates are free through Let's Encrypt and automatic on most modern hosts. If yours has expired, the underlying problem is usually that nothing is monitoring it.
Image alt text
Text describing what each image shows, read by screen readers and by anything that cannot see the picture.
This is an accessibility requirement first and a search benefit second. For a trade business it also happens to be where a lot of specific, useful description lives — a photo captioned "brick patio, Church Hill" says something a machine can use.
The fix: describe the image as you would
to someone on the phone. Leave alt empty for purely
decorative images rather than stuffing keywords into it.
Canonical URL
A tag naming the single official address for a page, so that the same content reachable several ways is not treated as several competing pages.
Sites reachable at both www and non-www, or
with tracking parameters appended, can end up splitting their own
signals between duplicates.
The fix: set a self-referencing canonical on every page and pick one hostname, redirecting the other to it.
Common questions
Is this the same as SEO?
It overlaps, but it is not the same thing. Traditional SEO works to earn a position in a list of links. These checks are about whether a machine can read, understand and quote your page — which is what decides whether an AI assistant names you in an answer. Structured data, answer-first writing and crawler access matter far more for that than keywords and backlinks do.
Will fixing these get me into ChatGPT's answers?
Nobody can promise that, and anyone who does is guessing. What these checks do is remove the reasons a machine would fail to understand or trust your page. That is a necessary condition, not a sufficient one.
Do I need llms.txt?
Not urgently. It is a young convention with no proven ranking effect. It is cheap to add, it removes ambiguity about what your business does, and 73.7% of Richmond business sites do not have one — so it is one of the few places where being early costs almost nothing.
How often should I re-check?
After any redesign, host migration or CMS update, and once a quarter otherwise. Most of these break silently — a plugin update rewrites robots.txt, a certificate lapses, a theme change drops the schema block. Nothing tells you.
Where do the Richmond figures come from?
A random sample of 2,500 active registered businesses across Richmond, Henrico, Chesterfield and Hanover, audited in August 2026. The methodology, confidence intervals and full dataset are published openly.