I shipped a round of checks to the Mega Analyzer yesterday based on patterns I keep seeing in real audits — small brand and copy issues that pass every SEO tool, look fine to a human reader, and quietly cost you Google visibility and donor or buyer trust. None of them are exotic. Most are 30-second fixes once you see them. The reason audit tools usually miss them is that they sit in the seam between SEO and copy — and most tools cover one or the other, not both.
If you run a site, here are the 12 patterns to look for. The Mega Analyzer now flags each one on any URL you paste in.
The "AI" typo that's actually "Al"
The most subtle one, and the one that made me write this round. Open the homepage of a few sites in the AI consulting space and search the H1 for "Power of AI." If the site uses a font like Inter or Helvetica, the rendered glyph for a lowercase L and a capital I are visually identical. So a headline that reads "Power of Al" (capital A, lowercase L) looks correct to the eye but contains a different sequence of Unicode codepoints — and Google's indexer, screen readers, Perplexity, Claude, and ChatGPT all parse the codepoints, not the rendered glyphs.
Try this: open a page in Chrome, hit Ctrl-F, click the "Aa" Match Case button, and search for Power of AI. Zero matches. Search for Power of Al. One match per H1 element. The typo can sit on a homepage indefinitely because every human proofreader sees what they expect to see.
Common confusable pairs in sans-serif fonts: capital I and lowercase l and the digit 1; uppercase O and the digit 0; capital B and the digit 8. The Mega Analyzer scans the title, H1, and H2 of every page for these patterns and flags any hit so you can verify in the DOM.
Dev-placeholder text leaking into production
Meta descriptions full of "lorem ipsum," "[enter X here]," "TBD," "your text here," or — and this is the one I see most often on smaller sites — non-English placeholders the developer used. Filipino "gwapo" (handsome), Spanish "guapo," generic "sample text." These end up in Google's index as the SERP snippet for the page. Most of them survive multiple redesigns because nobody on the editorial side speaks the language the developer was using as a placeholder.
The Mega Analyzer scans page title, meta description, every H1, and the first 4,000 characters of body text against a dozen common dev-placeholder patterns. One hit is enough to flag.
Joined-word headlines
You write a headline like "Let's Connect" and put a <br> between the words for the stacked-line look. The visible page reads fine. The DOM textContent reads LET'SCONNECT as one token. Screen readers announce one word. AI assistants quoting your H1 in an answer cite "LETSCONNECT."
Same pattern produces THANKYOU, WELCOMETO, SUCCESSFULPROJECTS. The Mega Analyzer flags two signatures: ALL-CAPS words with an apostrophe followed by 4+ more uppercase letters, and any single all-caps word longer than 13 characters (which almost always means two words mashed together, since "TECHNOLOGICALLY" and "CONSTITUTIONALLY" are the rare real-word exceptions).
The fix is a real space inside the heading, or two separate elements with CSS for visual stacking.
All-caps written into the HTML source instead of via CSS
Headings like <h1>WHO WE SERVE</h1>. The eye reads them as a stylistic choice. AI assistants quoting your H1 in an answer include the case literally — your heading appears as SHOUTY TEXT in citations. Worse, some screen-reader engines try to pronounce all-caps words letter by letter.
The convention is to write headlines in Title Case in the HTML and use CSS text-transform: uppercase for the visual treatment. The Mega Analyzer flags any H1 where more than 80% of letters are already uppercase in the source.
BlogPosting author declared as @type Person, but the name is the company
If you use a CMS that defaults the blog author to your organization's name but emits the author block with @type: Person, Google Search Console flags every post: "Invalid object type for field 'author'." The error sits in your GSC reports until someone notices.
The Mega Analyzer cross-references the author's @type against the name. If @type is Person and the name contains org-shaped tokens like LLC, Inc, Partners, Solutions, Corporation, Studio, Group, Holdings, Agency, or Consulting — it flags. Fix is either to switch the author to @type: Organization referencing the same @id as your sitewide Organization block, or — better for E-E-A-T — to set up real human author profiles in the CMS.
CMS-default author placeholders
The other variant of the same problem: author set to "admin," "administrator," "webmaster," "default," "wp-admin," "root." These are factory-default values that never got replaced. The Mega Analyzer flags any BlogPosting / Article with an author name matching a known CMS placeholder pattern.
Meta description over 170 characters
The existing Mega Analyzer flagged short meta descriptions (under 100 chars) but not over-long ones. Google generally truncates at around 155-160 characters, and descriptions over 170 are often rewritten entirely by Google — meaning you lose control of the snippet that appears in search results. Now flagged at 170+ with a recommendation to trim to 140-160.
og:image missing while og:title is present
A common pattern: the page has og:title and og:description set up (so it claims social-share readiness) but no og:image. The result is that LinkedIn, Slack, Facebook, and X shares all show a blank thumbnail. The Mega Analyzer used to check only for og:title; now it checks og:image separately.
twitter:card=summary with og:image present (mismatch)
When og:image exists and twitter:card is set to summary (the small thumbnail card), you almost certainly meant summary_large_image. The difference is a small icon-style preview versus a full-bleed banner. Mega Analyzer detects the mismatch and recommends the upgrade.
Funnel pages indexable by search engines
Thank-you pages, confirmation pages, "your meeting is booked" pages — these should be noindex because they have no value for someone arriving from a Google search and they actively compete with your real conversion-target page for the same query. The Mega Analyzer now detects URL patterns like /thank-you/, /ty-, /confirmation/, plus title patterns like "Thank You" or "Booked," and warns if the page is still indexable.
Thin pages still indexable and not canonicalized away
A page with under 1,500 characters of body text, no noindex, and no canonical pointing to a richer page. The Mega Analyzer flags this with three remediation options: expand the content, mark noindex, or canonical to a fuller page covering the same topic.
Missing preconnect / dns-prefetch hints when many third-party hosts are referenced
If a page loads resources from 5 or more external hostnames (Google Fonts, a CDN, a video stream, GTM, etc.) but has no <link rel="preconnect"> or <link rel="dns-prefetch"> hints in <head>, mobile LCP suffers by 200 to 500 milliseconds for no good reason. The Mega Analyzer counts unique third-party hostnames and flags absence of hints. Adding the top 3 to 4 hottest hosts is usually a one-line change.
How to run the checks yourself
Open jwatte.com/tools/mega-analyzer/, paste any URL, and the 12 new checks fire automatically alongside the existing 140+. No login, no install, no cost. The results card shows pass or fail for each check with a one-line explanation and a fix suggestion. There's a "copy AI fix prompt" button at the bottom that emits a single consolidated prompt you can paste into Claude or ChatGPT to get specific code suggestions for everything flagged.
If you want the broader self-help kit — installing Claude Code so it can apply fixes against your CMS API directly — the AI Terminal Kickstart walks through Windows and Mac setup in about 20 minutes.
For the deeper "why does this matter to AI assistants" context behind several of these checks, Your schema validates. Google AI Mode still can't find you covers the entity-anchor pattern. And the Nine AI Mode entity-binding bugs that pass every schema validator post covers the author-type mismatch specifically.
If you're building or running a small site and want the broader argument about keeping the whole stack affordable, The $50-a-Month AI Stack for SMB makes the case. That's also the context behind The $97 Launch — the under-$100 build-your-own-web playbook this whole free-tools site grew out of.
Fact-check notes and sources
- Lowercase
l(U+006C) and uppercaseI(U+0049) are different Unicode codepoints. Their visual identity in many sans-serif fonts is documented by the Unicode Consortium's confusables list at https://www.unicode.org/Public/security/latest/confusables.txt. - Google's SERP meta description truncation length: around 155-160 characters on desktop, less on mobile. Source: Google Search Central, "Control your title links and snippets in Google Search results."
twitter:card=summary_large_imagecard-size specification: https://developer.x.com/en/docs/x-for-websites/cards/overview/markup.- Google Search Console's "Invalid object type for field 'author'" error: documented at https://search.google.com/search-console/help under structured-data error reports.
<link rel="preconnect">LCP impact: web.dev's Largest Contentful Paint guide, https://web.dev/lcp/.- Mega Analyzer source and per-check logic: https://jwatte.com/tools/mega-analyzer/ — view source for the exact regex patterns and pass/fail conditions.
This post documents Mega Analyzer Round 10. The companion non-profit-aware round (Round 11) is described in 12 non-profit transparency checks the Mega Analyzer now runs.
