B2B Technical SEO

What this is

This is not a generic SEO retainer. It’s B2B technical SEO done by the team that builds your systems, so search is wired into your product and your funnel instead of running as a separate checklist someone reports on each month. The foundation underneath is the boring part done right: a site that search engines can crawl, understand, and rank, built that way from the first commit. It’s also what makes you eligible to be cited by AI, because AI search retrieves from the same indexes you’re trying to rank in. If the underlying index can’t find you, you’re not in the candidate set — ranking is the entry ticket, not a guarantee of citation.

Why the AI shift makes this matter more, not less

There’s a common assumption that AI search makes classic SEO obsolete. The plumbing says otherwise. The retrieval crawlers behind LLM answers mostly fetch your raw HTML and don’t run JavaScript — an analysis of 569 million GPTBot requests found AI crawlers effectively don’t execute client-side JS. So a single-page app that paints its content after hydration looks empty to them. That’s a build decision, not a content one: render the page server-side and the content is there in the first response.

The other half is what actually moves AI citations. The original GEO study (KDD ‘24) tested this directly and found that adding cited sources, statistics, and quotations to a page raised its visibility in generative answers by up to ~40%, with citing sources the single strongest lever. That maps cleanly onto what good editorial SEO already does. The shortcuts don’t hold up: SE Ranking’s study of nearly 300,000 domains found no correlation between llms.txt and AI citations (their model actually got more accurate once the file was dropped as a signal), and Google’s spam policies now explicitly name attempts to manipulate generative AI responses as spam. The durable play is the same as it ever was: a site that’s crawlable, well-structured, and worth citing.

What the audit covers

“Technical SEO audit” can mean almost anything, so here is what ours actually checks. Every item produces a pass or fail with evidence, and the output is a fix list ranked by impact — which we implement ourselves rather than handing over as a PDF.

• Raw-HTML render check. We curl your key pages and confirm the content is in the server’s response, not assembled later by JavaScript. AI retrieval crawlers read the raw HTML and move on, so a page that only exists after hydration is invisible to them. We’ve written up how to run this check yourself in two minutes if you want to self-check first.
• Core Web Vitals against the real thresholds. Google’s bar is LCP within 2.5 seconds, INP at or under 200 milliseconds, and CLS at or under 0.1, at the 75th percentile of real users — not a one-off lab run. Field data where it exists, Lighthouse where it doesn’t, and a per-page list of what fails and why.
• JSON-LD validation. Structured data only helps if it parses. We validate every block against the schema.org types and Google’s structured-data requirements, and flag markup that contradicts the visible page — which is worse than shipping none.
• hreflang reciprocity. Every language alternate has to point back at its counterpart; one-way pairs get ignored, and broken alternates can leave the wrong language ranking in the wrong market. We check each pair in both directions.
• Crawler access in robots and at the CDN. robots.txt is the easy half. The failure we see more often is edge rules — bot management, WAF defaults — silently blocking GPTBot, ClaudeBot, or PerplexityBot while the site looks fine in a browser. We test as the named crawlers and read the actual edge logs, not just the config.

None of this is exotic. Listing it means you can hold us to it — and most audits skip the first and last items, the two that now decide whether AI search can see you at all.

How we build it

We render server-side so the content exists in the HTML before any JS runs — on Astro for content sites, or Next.js App Router on OpenNext for Cloudflare when the app needs more. Either way the crawler, human or machine, gets a complete page on first request.

On top of that:

• Structured data. JSON-LD per Google’s structured-data docs, typed against schema.org, so an Organization, Article, Product, or FAQ is unambiguous to a parser instead of inferred from markup.
• Core Web Vitals as a budget, not an afterthought. We hold the page to the web.dev Core Web Vitals thresholds — LCP, INP, CLS — and treat a regression as a build failure, because slow pages get crawled less and rank worse.
• Architecture and internal linking mapped to the queries that matter, so crawl depth and link equity land on the pages you want found.
• Measurement wired through GA4 so referrals from AI assistants and organic search are visible, not guessed at.

Where this is heading

Buyer research is moving toward AI intermediaries — more B2B purchasing now runs through systems that read the web and decide what to surface. Whether a person or an agent is doing the searching, the requirement is the same: a fast, crawlable, well-structured site that a retrieval system can read and trust. We build that foundation once, then maintain it as the indexes and crawlers change.