Nothing is blocking Google. Robots, canonicals, sitemap and status codes are all correct. Google is simply crawling slowly because every page it fetches is a megabyte or more, most of that content is repeated five times inside each page, and the site is new enough that it has not earned a bigger crawl budget yet.
Cut the page weight and link the articles from the homepage, and the rest of the site gets indexed on its own. The favicon will follow the homepage recrawl.
Nothing to fix
Checked: the homepage declares five icon tags; every file loads with HTTP 200 — favicon.ico (16, 32, 48 and 64 px inside), favicon-96.png (96×96), favicon-192.png (192×192), favicon.svg, and the Apple touch icon. Square, correct sizes, crawlable.
Google’s own rule: the icon appears after Google recrawls the homepage, which “can take anywhere from several days to several weeks.” Your homepage is 5.75 MB and the site is ten weeks old, so it is at the slow end. Optional nudge: keep one primary tag — <link rel="icon" href="/brand/favicon-192.png" sizes="192x192"> — and use URL Inspection → Request indexing on the homepage.
Why only 16 pages are in Google
- The site is ten weeks old.Domain registered June 13, 2026. New sites get a small crawl budget until Google sees demand for them. Not a fault — a starting condition that the next four points make worse.
- Every page is 1–6 MB of HTML.Homepage 5,751,878 bytes; About 2.5 MB; Pricing 2.05 MB; each blog post ~1.84 MB; median page 1.03 MB. Googlebot spends its budget in bytes and time. Fifty of your pages cost what five thousand normal pages would.
- Each page contains five copies of itself.The site is a Framer design exported and re-hosted on Vercel. The export renders every screen-size breakpoint into the file: the H1 “Pharmacogenomic AI” appears 5 times, the hero video is embedded 5 times, and 1,838 hidden breakpoint containers sit in the homepage. Google sees the duplication and discounts the page.
- The 23 Insights articles are only reachable from the Insights index.They are the lightest, most indexable pages on the site (about 22 KB each, 1,600 words, cited to CPIC and FDA) and the homepage does not link to a single one. Pages Google can reach only through a sitemap get crawled last.
- The homepage links to dead ends.It links 15 times to
/casesand 10 times to/career(both just redirect) and 5 times to/404— a “page not found” page that returns HTTP 200. A soft-404 linked from the homepage is a quality signal in the wrong direction.
What is inside the 5.75 MB homepage
| Component | Bytes | Note |
|---|---|---|
| Inline SVG (431 graphics) | 1,749,876 | Should be a handful of image files, cached once |
| Inline CSS | 1,162,536 | One stylesheet, cached, would be under 100 KB |
| Everything else in the HTML | 2,839,466 | 12,109 <div>s; five breakpoint copies |
| Hero video (1.1 MB) × 5 downloads | 5,572,573 | Same file fetched five times per visit |
| Second autoplay video | 3,020,960 | Autoplays on load |
| Largest Framer JS chunk | 1,846,925 | Plus 58 more module files |
| PNG images (each) | 521,497–674,778 | WebP at the same quality: under 100 KB |
Result on Google’s PageSpeed Insights (mobile): performance 33, first paint 7.3 s, largest paint 19.8 s, 1,050 ms blocking. The desktop run timed out on Google’s side; run locally it finishes at 55 with a 3.3 s largest paint. Server response is 20 ms — Vercel is not the problem, the payload is.
Things a reviewer notices
| Item | What we found | Status |
|---|---|---|
| Template copy left in | “Identifying high-impact AI opportunities across operations, workflows, and data processes…” appears 15 times, rendered at 10 px. It is filler from an AI-agency Framer template; a Google search for the sentence finds exactly one page on the web — yours. | remove |
| Social icons | Link to facebook.com (×10), x.com (×9), pinterest.com (×8), instagram.com (×8) — the networks’ homepages, not your profiles. | fix |
| blog.signalpgx.com | Hosted on a third-party AI blog service; footer reads “© 2026 Tarek Younis’s Organization” (the service’s placeholder); first request returned a 503; 0 of 31 posts indexed. | decide |
| “Founded: 2024” | Stated in llms.txt and the Wikidata entry; the domain was registered June 13, 2026. | align |
| Publishing pattern | 51 of 55 pages last-modified Aug 22; 23 articles all dated early August; Wikidata entry created Aug 12; press release Aug 17. Nothing wrong with any one of these; together they look like a burst, and burst-shaped new sites get held back. | pace |
What to do, in the order it pays off
- Publish from Framer directly, or do a proper build.Framer’s own hosting serves one breakpoint and optimises assets; the scraped export does neither. This single change fixes causes 2 and 3 and most of the page-weight table.
- Drop the autoplay videos, or keep one lazy-loaded copy.8.6 MB of the 15 MB load is video on a page whose reader is a lab director.
- Link the Insights articles from the homepage and navigation.They are the pages most likely to be indexed and to bring the right visitors.
- Remove the dead links and the template copy; point the social icons at real profiles.
- Move the blog under signalpgx.com/blog, or retire the subdomain.Thirty-one posts that Google has not indexed are not helping anywhere.
- Then: submit the sitemap in Search Console and request indexing on the ten most important URLs.Expect weeks, not days.
Five questions that shape the fix
Answer what you can now; skip the rest. When you press Send answers, your reply goes straight to us.