AI Overviews: is your CMS ready for generative search?

For Google, there is nothing special to do to appear in AI Overviews. According to its documentation, the SEO fundamentals are what count. What your CMS changes is your ability to maintain those fundamentals across thousands of pages, in several languages and over time.

Outside Google, only one task is genuinely specific: letting the search crawlers of ChatGPT, Claude and Perplexity read your pages. So the real question is not whether your CMS "does AI", but whether it lets you publish pages that are indexable, readable and current.

Do AI Overviews require any special optimization?

No. On its "AI features and your website" page, Google states that there are no additional requirements and no special optimizations needed to appear in AI Overviews or AI Mode. There is no special schema.org markup to add either.

To be shown as a link in an AI Overview, a page has to be indexed and eligible to appear in Search with a snippet, as with any Google result.

Its guide to generative AI features, updated in July 2026, goes further. From Google’s perspective, optimizing for generative AI search is optimizing for search, which means it is still SEO. It lists llms.txt files, "chunking" content into small pieces and rewriting text "for AI" among the things you can ignore for Google Search.

What does your CMS actually change?

It changes your ability to maintain these fundamentals at scale. Across thousands of pages in several languages, updated by several teams, whatever is not automated eventually drifts. SEO quality then depends on implementation more than on the architecture you chose. Six points are worth checking on your platform.

• Server-rendered HTML. Google can execute JavaScript, but it reminds site owners that server-side rendering or pre-rendering is still a good idea because not all bots can run JavaScript. Text that only appears once a script has run may therefore be missed by some of them.
• Structured data generated by templates. Article, Organization or BreadcrumbList markup is not required for AI Overviews. What Google does ask is that structured data matches the visible text on the page. Generated by the template, it stays accurate with every update.
• A title, a meta description and hreflang tags for each language. Each language version carries its own metadata and declares its translations, so Google can point each reader to the right version.
• Fast updates and accurate dates. Correcting a figure or a recommendation should take minutes, review included. The displayed date should reflect what actually changed.
• Page-level control over indexing. Your robots.txt, your sitemap and the noindex and nosnippet rules decide what Google can crawl, index and reuse, AI Overviews included. These settings should be within reach of the teams who publish.
• Performance. Google lists page experience, correct display on every device and reduced latency among the fundamentals that still apply to generative AI search.

A headless architecture is not a prerequisite for appearing in these answers. Its purpose is to reuse content across several channels. For search engines, what counts is the published page their crawlers can read.

What should you do for ChatGPT, Claude and Perplexity?

Let their search crawlers access your pages. This is the only work that is genuinely specific to these engines, because they crawl the web with their own bots instead of relying on Google’s index. Each vendor documents several crawlers that should not be confused.

• Search crawlers feed answers that include links. They are OAI-SearchBot for ChatGPT, Claude-SearchBot for Claude and PerplexityBot for Perplexity. OpenAI states that sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers, apart from navigational links.
• Training crawlers collect content to train models. They are GPTBot at OpenAI and ClaudeBot at Anthropic. On Google’s side, Google-Extended is not a separate crawler but a control token. It governs the training of Gemini models and the grounding of answers in Gemini Apps. According to Google, it does not affect a site’s inclusion in Google Search.
• User-triggered visits are a third case. ChatGPT-User, Claude-User or Perplexity-User read a page when someone asks a question. OpenAI notes that robots.txt rules may not apply to ChatGPT-User. Perplexity says Perplexity-User generally ignores them.

Blocking a training crawler while allowing a search crawler is therefore possible. OpenAI explicitly documents that the two settings are independent.

That leaves the firewall. An open robots.txt is not enough if a web application firewall or a CDN blocks these crawlers upstream. Google lists allowing crawling by any CDN or hosting infrastructure among the fundamentals. Check your server logs to confirm these crawlers actually receive your pages.

Can you guarantee being cited by AI?

No. No page structure, markup or file guarantees a citation. Google itself notes that following every guideline does not guarantee that a page will be crawled, indexed or served.

What you can do is make a page readable and quotable. Answer the question in the first sentences of each section. Write passages that make sense on their own. Offer information or a perspective readers cannot find elsewhere. Keep figures sourced and dated. These practices serve the reader first.

How do you measure the impact of AI Overviews on your traffic?

By tracking two indicators: the impressions your pages get in Google’s generative answers and the click-through rate of your most exposed pages. The Search Console Generative AI performance report covers AI Overviews and AI Mode. Google says the report has been available to all websites worldwide since August 31, 2026. It shows impressions by page, country, device and date, with no clicks or queries.

On clicks, a Pew Research Center study of the actual browsing of 900 U.S. adults in March 2025 gives an order of magnitude. Of the Google searches studied, 18% produced an AI-generated summary and 58% of respondents saw at least one during the month. With a summary, users clicked a traditional result link in 8% of visits, compared with 15% without one. Links inside the summary were clicked in just 1% of visits. Browsing sessions ended on 26% of results pages with a summary, compared with 16% of pages without one.

The same study found that longer searches, or searches phrased as questions, produce a summary more often (53% of searches of ten words or more, compared with 8% of one- or two-word searches). For these queries, track the click-through rate of the pages concerned in the standard performance report, not just their position.

Which questions should you ask your team or your CMS vendor?

• Do our pages deliver their main text in the HTML sent by the server?
• Is our structured data generated by templates and consistent with the visible text?
• Does each language have its own title, meta description, hreflang tags and sitemap?
• Can we correct and republish a page in minutes, with an accurate update date?
• Does our firewall or CDN let through the search crawlers we allow?
• Do we track generative impressions and the click-through rate of our most exposed pages?

How does Jahia handle these fundamentals?

At Jahia, these fundamentals are part of the CMS and documented on the Academy. The robots.txt file is edited from the interface. The sitemap module generates one file per language automatically, taking noindex pages into account. Readable URLs (vanity URLs) are managed page by page and hreflang and canonical tags are generated automatically.

For rendering, Jahia applies an approach that favors server-side rendering to its JavaScript modules. Publication workflows let a reviewer validate an update before it goes live.

These capabilities help maintain the fundamentals across thousands of pages and several languages, which is what search requires, generative or not. For the role of AI inside the CMS itself, see our page on the enterprise CMS with built-in AI.