The honest answer
We ran the same 12-point SEO audit on 50 digital agency websites between February and April 2026. The methodology was not automated. We used PageSpeed Insights for Core Web Vitals data, Screaming Frog for crawl analysis and canonicals, Google's Rich Results Test for structured data validation, and manual inspection for Search Console status. Every site got identical treatment.
Forty-seven of fifty had at least three critical failures. Three had all twelve. None passed clean.
These are agencies actively selling digital marketing and SEO services to clients. Most of their own sites are doing the opposite of what they tell clients to do. That is not a critique of their client work. It is data about a specific organizational blind spot: the cobbler's children problem, applied to search marketing.
What the Audit Covered
The 12-point checklist maps directly to items Google has documented in its Search Central guidance and Core Web Vitals framework. These are not obscure ranking theories. They are documented webmaster requirements that Google has been explicit about for years.
The twelve checks:
- Search Console configured, verified, and free of manual actions
- XML sitemap submitted to Search Console
- Organization schema implemented without validation errors
- Service or Product schema on primary service pages
- FAQPage schema on any FAQ content sections
- LCP (Largest Contentful Paint) under 2.5 seconds on mobile
- CLS (Cumulative Layout Shift) under 0.1 on mobile
- INP (Interaction to Next Paint) under 200ms
- Canonical tags present and consistent across key pages
- Image alt text on all meaningful images
- Meta descriptions between 130 and 160 characters on homepage and service pages
- Homepage links to primary service pages with descriptive anchor text
We chose these 12 because they represent the foundation Google explicitly tells webmasters to get right first (per Google Search Central documentation, 2024). Getting them wrong does not mean a site will never rank. It means the site is ranking despite problems, not because of a sound technical foundation.
The 12 Findings, Ranked by Frequency
Here is the complete dataset from 50 audited agency sites:
| Finding | Sites Affected | Percentage | Avg Severity | Easy Fix? |
|---|---|---|---|---|
| Missing or broken structured data | 47 | 94% | 4/5 | Medium |
| LCP above 2.5s on mobile | 39 | 78% | 4/5 | Medium |
| No Search Console configured | 38 | 76% | 5/5 | Easy |
| Meta descriptions over 160 chars or missing | 35 | 70% | 2/5 | Easy |
| No XML sitemap submitted | 33 | 66% | 3/5 | Easy |
| Thin content (no pages above 800 words) | 31 | 62% | 3/5 | Hard |
| Missing image alt text on majority of images | 29 | 58% | 2/5 | Easy |
| Duplicate or missing canonical tags | 26 | 52% | 4/5 | Medium |
| CLS above 0.1 | 24 | 48% | 3/5 | Medium |
| Weak internal linking from homepage | 22 | 44% | 3/5 | Medium |
| INP above 200ms | 18 | 36% | 3/5 | Hard |
| Mobile usability test failures | 9 | 18% | 4/5 | Medium |
Structured data: the 94% problem
This number surprised us even though we expected it to be high. FAQPage schema alone can produce rich results that double a listing's visual footprint in the SERP. Organization and Service schema help with entity recognition in AI Overviews and Knowledge Graph entries. Google has documented all of this clearly (per Google Search Central, 2024).
Of the 47 sites with structured data failures: 31 had no structured data at all. Sixteen had schema with validation errors caught by the Rich Results Test. Two of those 16 had schema that looked complete on the surface but used incorrect property types, making the markup non-functional. We found one agency with FAQPage schema where every answer property was an empty string. That is worse than no schema because it signals an attempt at manipulation without substance.
No Search Console: the 76% problem
Three out of four agencies selling digital marketing services cannot see their own search performance data. No visibility into what queries are driving clicks. No alerts when Google flags crawl errors. No way to detect index coverage gaps before they cost months of ranking.
During our manual review, four sites had pages completely missing from the index. None of the agencies appeared to be aware of this. One had their entire blog directory blocked by a robots.txt disallow rule that a developer had added for a local development environment and never removed before pushing to production. That blog had 140 posts. Zero of them were indexed.
LCP: the technology correlation
The 78% LCP failure rate is not random across technology stacks. It concentrates on specific platform choices. Sixty-two percent of the slow sites were running WordPress without meaningful caching or image optimization. Seven were on Squarespace. The fastest sites were on Next.js, Nuxt, or Astro with CDN-delivered images.
LCP above 4 seconds appeared on 23 sites. LCP above 6 seconds appeared on 8, all WordPress. One site loaded a 3.4MB hero image on mobile with no compression and no lazy loading. That is not a Core Web Vitals score problem. That is a workflow problem.
What the Top Three Sites Did Differently
Three agencies passed eight or more of the 12 checks. They shared three observable characteristics.
They treated their site as a product under active development, not a brochure. Each had blog content over 1500 words on at least five pages. Structured data covered every service page. Content was updated on a schedule, not pushed out and forgotten. One published a detailed case study every six weeks with specific traffic and conversion metrics included. That is the kind of first-hand experience documentation that Google's Helpful Content guidance explicitly describes as a quality signal.
They had a named person responsible for SEO maintenance. Not a contractor who handled it once a year. An internal person who ran monthly crawls, reviewed Search Console weekly, and kept structured data current as the agency's services evolved. Technical SEO health is maintenance work, not a one-time project. The data shows what happens when it gets treated as a one-time project.
Two of the three published their own performance data. One had a public page showing 24 months of organic traffic trends alongside their Lighthouse scores. Unusual, honest content attracts links. That site had 34 referring domains pointing to their transparency page alone. (And yes, that matches what first-hand experience research at SparkToro has found: content that exposes real internal data earns links that promotional content does not.)
The pattern is consistent: agencies that apply their own advice to their own site outperform agencies that do not. That sounds obvious. The data says it is not obvious enough to act on.
The Finding That Surprised Us
We expected the structured data failures. We expected Lighthouse scores below 50. What surprised us was the internal linking gap: 44% of agency homepages had no descriptive anchor text links to their primary service pages.
Homepages linked to About, Contact, and Book a Call. They did not link to "SEO Services for SaaS Companies" or "Next.js Development Agency." This is a PageRank distribution failure. The homepage typically carries the highest link authority of any page on the domain. Failing to pass that authority to service pages through descriptive internal links leaves a ranking signal unused that costs nothing to fix.
We ran an informal before-and-after test on one consenting agency with exactly this gap. They added six internal links from the homepage to service pages using keyword-rich anchor text. Fourteen weeks later, four of the six target pages had moved from page 3 to page 1 for their primary keywords. One data point. Not a controlled experiment. But consistent with what PageRank theory predicts and with what Google's Search Central documentation says about internal link equity.
The content thinness finding was a close second for surprising us. Sixty-two percent of audited sites had no page over 800 words. Not one. Several had blogs consisting entirely of 150-word news items and 250-word project announcements. This is not content that earns search visibility. It demonstrates a misunderstanding of what earns search visibility. Google has been explicit in its guidance: helpful content requires depth, first-hand expertise, and content that serves readers rather than search engines. A 200-word blog post demonstrates none of those qualities.
The contrarian take
There is an argument that agencies are busy serving clients and do not have bandwidth for their own site maintenance. That argument explains the situation but does not excuse it. An agency that cannot maintain its own digital presence is making an implicit claim about its priorities that prospects will notice. If your potential clients are running any kind of due diligence, they are looking at your site's technical health. A 6-second LCP and zero structured data on a site that sells SEO services is a trust-eroding signal.
The three top performers had no more resources than the worst performers. They had a different allocation of priorities.
What this means in practice
Run your own site through this same 12-point checklist before the end of this week.
Start with the easy fixes: Search Console setup, XML sitemap submission, image alt text, and meta description audit. These take under two hours and remove obstacles Google uses to deprioritize pages. Then add structured data. A developer who knows what they are doing can implement Organization, Service, and FAQPage schema in one day. Then fix internal linking from the homepage.
LCP improvements are harder. If your site is on WordPress without a caching layer, the fastest path is moving to a static framework or adding a CDN and compressing images above the fold. If you want a full audit against this same checklist delivered with a prioritized fix list, our web development and SEO services use exactly this methodology.
The data is clear: 47 of 50 agencies had correctable problems. Most of those problems were on the easy list. Fix your own house first.
