Workstream · Technical SEO
Technical SEO: fix what blocks everything else
Crawling, indexing, JavaScript rendering, performance. As long as crawlers stumble on your site, the content you produce and the links you earn count for nothing. That now extends to AI crawlers: if GPTBot and PerplexityBot cannot read your pages, no AI answer will cite you. We fix it in your code, not in a recommendations deck.
Technical SEO, in one sentence
Technical SEO is the set of interventions that make a website crawlable, indexable and fast, for classic search engines and for generative engines such as ChatGPT alike. It produces no content and earns no links. It conditions what content and links can produce next, on Google and inside AI answers.
A URL Google never crawls ranks for nothing. A page an engine cannot render cannot be cited. Technical SEO is one of the six workstreams of organic search, and it is the only one whose absence cancels the other five. Content, links and authority all depend on a site crawlers can actually read.
The same dependency now applies to AI answers. ChatGPT, Perplexity and Google AI assemble their responses from pages their crawlers fetched and read. If your pages fail at that first step, no editorial effort downstream will place you in an answer. Fixing that access is technical work, and it happens in your code.
The signals that bring you here
Strategic URLs crawled by Google but absent from the index. They cannot bring in a single visit.
A JavaScript site whose rendered version contains no usable text or links. Crawlers leave with nothing.
Third-party scripts or a heavy theme dragging LCP and CLS down on mobile.
Successive hops between URLs that stretch every crawl path and waste crawl budget.
GPTBot or PerplexityBot blocked in robots.txt, or served pages whose content only exists after hydration.
Five signals send teams to this page: strategic URLs crawled but not indexed, JavaScript pages whose rendered version holds no usable text, Core Web Vitals sinking on mobile, redirect chains between internal URLs, and AI crawlers locked out of the site. Every one falls inside the scope of the Graine audit.
The three barriers that keep AI crawlers out
Three technical barriers come up again and again when AI engines try to read a website: heavy JavaScript, a missing llms.txt file, and incomplete Schema.org markup. Only 3.2% of websites have deployed llms.txt. Each barrier has a defined correction, and each is applied in your code, not in a memo.
| Barrier | What the crawler gets | The correction | Horizon de résultat |
|---|---|---|---|
| Heavy JavaScript | GPTBot and PerplexityBot fetch raw HTML and do not run your scripts. Content that appears only after hydration never reaches them. | Server-side rendering or pre-rendering, so the full text and links exist in the initial HTML response. | |
| Missing llms.txt | No curated entry point for AI systems. The crawler is left to guess which URLs represent you. | An llms.txt file at the root, listing your key pages in plain text with one line of context each. | |
| Incomplete Schema.org | An ambiguous entity. The engine cannot state with confidence who you are, what you sell, or where. | JSON-LD written template by template and validated: Organization, Service, FAQ, the types your pages actually need. |
How many websites have an llms.txt file?
3.2% of websites have deployed llms.txt, the plain-text file that hands AI systems a curated map of your site. Adoption is early. That is the point: deploying it now costs little and removes ambiguity about which pages should represent you when an engine builds an answer.
llms.txt is a plain-text file placed at the root of a website that lists the pages an AI system should read first, with a short description of each one. It does for AI crawlers what sitemap.xml does for classic ones: a declared map instead of a guessed one.
No engine publicly commits to obeying llms.txt today, and we say so plainly. We deploy it because the cost is low, the ambiguity it removes is real, and it is no substitute for the two corrections that matter more: content present in the initial HTML, and structured data that names your entity.
We write it during the Structure phase, alongside your JSON-LD and your internal linking. One file, your key pages, one line of context each, maintained as the site evolves.
What we do, concretely
What crawlers really fetch on your site, URL by URL: Googlebot, GPTBot, PerplexityBot. And the share of crawl budget spent on URLs with no commercial value.
We settle what should exist for engines: consolidation, canonicals, robots directives, URL parameters.
Server rendering, hydration, links crawlers can actually follow. The audit reads the rendered DOM, not the source file.
robots.txt reviewed for AI user agents, llms.txt written and deployed, key content present in the initial HTML response.
Core Web Vitals measured from your real visitors, CrUX data in hand, then corrected in the code.
JSON-LD written template by template and validated, so each page states explicitly what it is about and which entity stands behind it.
An alert whenever indexing or performance drops after a release.
Seven areas make up the technical SEO service: log analysis, indexing decisions, JavaScript rendering, AI crawler access, field performance, structured data, and regression monitoring. Every one ends in a correction shipped in your code, then re-measured on the same URLs. Nothing stops at the recommendation stage.
How we work
The work runs in three phases: measure, fix, structure. We start from what crawlers actually do on your site, server logs first, before touching anything. Corrections are then written in your code and verified against the same indicators, on the same URLs, that the first measurement recorded.
- 01MeasureWe record what crawlers actually do on your site before any correction: server logs and crawl budget, 404 errors, redirect chains and loops, robots.txt and sitemap.xml reconciled. AI user agents are read in the same logs.Deliverable : A prioritized list of corrections, each one sized in days.
- 02FixCorrections are written in your code, alongside your developers. CSS, JavaScript and images reworked. LCP and CLS tracked on field data. Cache and CDN configured for your hosting setup.Deliverable : The same indicators re-measured on the same URLs, before and after.
- 03StructureA readable site structure and explicit markup give engines what they need to identify each page's subject. Heading hierarchy reviewed, JSON-LD set per template and validated, llms.txt deployed, internal linking redistributed toward the pages that sell.Deliverable : Pages that state what they are about, to Google and to AI engines alike.
What an audit finding looks like
Every finding in the Graine report (Graine is our diagnostic tier, from €2,670) arrives measured, graded and paired with its fix. The two lines below illustrate the report template. They are examples, not client data: your own values come out of your own audit.
| Finding | Severity | What was measured | The fix |
|---|---|---|---|
| Pages not indexed | Critical | 81% of strategic URLs crawled by Google but absent from the index | Log analysis, crawling redirected toward business pages, the rest consolidated |
| Redirect chains | Moderate | 1 internal URL in 6 passes through two or more hops before resolving | Internal links rewritten to the final URL, redirect table flattened |
Technical work we have already shipped
Pennylane grew SEO traffic by 118% after a program combining technical optimization, educational content and semantic targeting, with qualified leads up 41% and conversion to product demos up 27%. Alma rebuilt its editorial and technical acquisition strategy in 3 months. Both are published as case studies.
Neither program was technical work alone, and we do not present it otherwise. Content and semantic targeting ran in parallel with the technical corrections. What was measured, what was corrected and over which period: the detail sits in each published case study.
Why technical SEO decides your AI visibility
AI queries grew 494% between 2023 and 2026, and 48% of French people now use a generative AI. The answers those tools give are assembled from pages their crawlers could fetch and read. Technical SEO is the discipline that makes your pages part of that raw material.
Selection is tightening. 38% of the sources in Google's top 10 are cited in AI Overviews in 2026, down from 76% in 2025. Engines quote fewer pages, and the pages they keep share traits technical work produces: complete rendering, explicit markup, fast delivery, an unambiguous entity.
Classic search still dominates raw volume: Google records 84 billion visits a month, versus 5.5 billion for ChatGPT. Good technical SEO never makes you choose. One codebase serves both audiences, because the engines share plumbing: ChatGPT leans on its corpus and Bing search, Gemini on the Google index and its SEO signals. A site that crawls clean for one crawls clean for the others.
This page covers the technical discipline. How we measure AI visibility (the Oracle audit, your state per engine as Absent, Mentioned or Recommended) and what each program costs are documented on our Generative Engine Optimization agency page and in our guide to what a GEO agency costs. Method and prices live there, in one place, versioned.
Is technical SEO ever fixed once and for all?
No. Every release can reintroduce a regression: a forgotten noindex tag, a third-party script that drags down rendering, a redirect added on top of another. That is why regression monitoring is part of the workstream, with an alert whenever indexing or performance drops after a deploy.
How long before you see an effect?
An indexing fix shows up when the affected URLs are recrawled. We publish no average timeframe: it depends on your site's crawl frequency, which the audit measures at the start. For AI visibility inside an ongoing program, first citations typically move in 6 to 10 weeks.
Do you fix it, or do we?
Both options exist. With the Graine audit (our diagnostic, from €2,670), you leave with a prioritized list, every correction sized in days, that your team can execute. With Pousse (our growth program), our consultants work in your code, alongside your developers. Every consultant is senior. Your file is never handed to a junior.
What does technical SEO change for AI engines?
A generative engine can only reuse what it could fetch and read. Complete rendering, accessible pages, explicit markup: the same corrections serve classic crawling and citation by ChatGPT, Perplexity or Google AI.
Is llms.txt part of the service?
Yes. Only 3.2% of websites have deployed llms.txt, and we treat it as one of the three technical barriers to AI access, next to heavy JavaScript and incomplete Schema.org. It is written and deployed during the Structure phase.
What does a technical SEO audit cost?
€2,670 excl. VAT, as part of the Graine audit: at least 3 days of analysis, delivered in 3 weeks, covering technical health, content, authority and your visibility in AI engines. If you continue into a program within 60 days, the amount is deducted from your first month. The full grid, every program included, is in our guide to what a GEO agency costs.
Start by finding out what blocks you
The Graine audit covers technical health, content, authority and your visibility in AI engines, for a published price of €2,670 excl. VAT: at least 3 days of analysis, delivered in 3 weeks. You get a written report, every correction sized in days, and a prioritized roadmap.
If you continue into a program within 60 days, the audit amount is deducted from your first month. And if you want to talk it through first, the first 30 minutes with a consultant are free, with no obligation.
Not sure the problem is technical? Run the free scan. It checks whether AI engines recommend you and returns a verdict in 60 seconds, no account, no card. If the verdict is Absent, this page is usually where the fix starts.
Case studies, measured results
How Pennylane strengthened its SEO presence on key accounting and financial queries
With Botanik, Pennylane structured a comprehensive SEO strategy combining technical optimization, educational content, and semantic targeting to strengthen its position in the financial management tools ecosystem.
Franck Neuenschwander · founder, author of the Oracle method
The method you just read, Organic Inference™, is mine: published, versioned, with its limits in writing. Fourteen years on the advertiser side, and the only agency I have ever worked at is the one I run: if you want to discuss it, you discuss it with me, not with a salesperson.
Thirty minutes with a consultant
Not a salesperson. We tell you which lever fits your case, and if none does, we say that too.
Are you absent, mentioned or recommended?