SaaS Technical SEO for AI Products That Ship Every Week

Sitemaps get pages discovered. They don't make pages worth citing. ChatGPT search and Copilot draw on Bing's index, so Bing coverage matters, but answers also need pages that render without JavaScript, state plainly what you do and are backed up by other sites. We check each of those layers, starting with what the crawlers actually fetched.

Either can work. What matters more is that docs link back to the product pages they support, carry correct canonicals across versions and don't compete for the same commercial terms. If you are already on a subdomain and it is healthy, we usually leave it alone. A move is only worth the risk when the current setup is clearly costing you.

For most startups, allow the crawlers that feed AI search results, such as OAI-SearchBot, PerplexityBot and Anthropic's search crawler, because blocking them can keep your pages out of those answers. Training crawlers such as GPTBot and ClaudeBot are a separate business decision. We lay out the trade-off, set the robots rules you choose and confirm your CDN and firewall respect them.

It can be. If the copy only appears after JavaScript runs, some crawlers see a near-empty page. Google renders JavaScript eventually, but many AI crawlers read the raw HTML and move on. We test what each crawler receives, then recommend server-side rendering, static generation or prerendering for the pages that matter, not a full rebuild.

Some changes register as soon as engines recrawl, such as unblocking a crawler or correcting a canonical. Index cleanup on a large site is slower, because engines drop and re-evaluate pages at their own pace, and rankings settle over weeks to months after that. We record the baseline first so each change can be traced rather than assumed.

Both, depending on access. Where we can work in the CMS, the robots file, redirects or schema directly, we make the change ourselves. For framework or infrastructure work, your engineers get a ticket with the exact fix, the affected URLs and how we will verify it. Then we check the release, not just close the ticket.

Usually Organization, SoftwareApplication or Product, FAQ where you have real questions, and breadcrumbs. The markup matters less than consistency. Your schema, homepage copy, pricing page and third-party profiles should state the same category, name and core features, so engines are not left reconciling facts that disagree with each other.

The site carried a large number of pages that earned nothing and diluted the ones that did. We pruned and consolidated them, rebuilt the landing pages that mattered and kept the index focused on pages worth ranking. Keywords in Google's top three grew 362%, and pages cited in AI answers went from 4 to 279.

It depends on the size of the site and whether we are clearing a backlog once or covering every release. We don't publish prices. After a free consultation we send a plan that lists the technical work in priority order, with a quote, and the work starts as soon as you approve it.