# Google Lighthouse Adds Agentic Browsing Checks and Cloudflare Adds AI Traffic Controls > Source: > Published: 2026-07-30 20:00:00+00:00 ### Gemini 3.6 Flash and 3.5 Flash-Lite Ignore Temperature, Top-P, and Top-K Google's Gemini 3.6 Flash and Gemini 3.5 Flash-Lite ignore temperature, top_p, and top_k, according to Google's current migration guide. Existing Gemini requests… Google Lighthouse and Cloudflare are adding different controls for the same emerging class of automated web visitors. Lighthouse added an experimental Agentic Browsing category. Before Lighthouse 13.3, Lighthouse did not include the Agentic Browsing category in its default configuration. The change gives developers and site owners a new diagnostic, while Cloudflare's current and planned controls require them to decide which automated visitors their sites support. The two products do not establish a shared policy. Lighthouse measures whether a page can be used by an automated client; Cloudflare lets a site operator control access by type of automated traffic. The useful comparison is therefore narrow: machine-readable design and traffic-access policy are separate choices. [Lighthouse's documentation](https://developer.chrome.com/docs/lighthouse/agentic-browsing/scoring) calls the category experimental and says testing requires Chrome 150 or later. The category runs deterministic audits: fixed checks intended to be reproducible in a CI/CD pipeline. It shows a pass ratio and individual pass-or-fail audit results, rather than a weighted 0–100 score or a definitive ranking. The checks cover WebMCP, a browser API for exposing a site's tools and forms to agents; the accessibility tree, a structured map of page elements that an agent can use; cumulative layout shift; and `llms.txt` . The accessibility-tree checks cover programmatic names, valid roles and relationships, and interactive content that is not hidden from that tree. The WebMCP audit watches Chrome DevTools Protocol events for declarative and JavaScript tool registration. The [Lighthouse 13.3 release notes](https://github.com/GoogleChrome/lighthouse/releases/tag/v13.3.0) say that the agentic-browsing category was added to the default configuration and was expected to ship in Chrome 150 DevTools and PageSpeed Insights. That makes the category a way to inspect particular implementation details; it is not a Google Search ranking signal. Cloudflare says [Browser Integrity Check](https://developers.cloudflare.com/waf/tools/browser-integrity-check/) is enabled by default. Browser Integrity Check is a Cloudflare feature that challenges suspicious HTTP clients: its documentation says it looks for headers commonly abused by spammers and can deny access or challenge clients with no user agent or a non-standard user agent. Site operators can disable it globally or selectively with a configuration rule or a custom-rule skip action. [Bot Fight Mode](https://developers.cloudflare.com/bots/get-started/bot-fight-mode/) is a separate optional free product. It identifies traffic matching known-bot patterns and issues CPU-intensive challenges. Cloudflare warns that it can challenge API or mobile-app traffic, and says it cannot be bypassed with WAF custom rules or Page Rules; an operator that needs exceptions must use a different product or configuration. Cloudflare's July 1, 2026 [AI-traffic announcement](https://blog.cloudflare.com/content-independence-day-ai-options/) introduced settings for Search, Agent, and Training crawlers. In that taxonomy, an Agent is a user-directed agent visiting a page on behalf of a human. Search is crawling that helps a site appear in search results; Training is crawling to train or fine-tune models. The announcement says that on September 15, 2026, new domains onboarding to Cloudflare will block Training and Agent by default on pages that display ads. Search will remain allowed by default. Those are planned, configurable defaults for new domains, not a statement that every existing Cloudflare customer is blocking agent traffic today. Cloudflare also says multi-purpose crawlers will be subject to the most restrictive applicable rule. The documented products support a limited inference: a site can improve the structure an agent encounters while independently choosing whether that agent's traffic is allowed, challenged, or blocked. That is a configuration decision for the operator, not a conclusion in either company's documentation about which automated visitors every site should admit. Give Vroni a GitHub issue, bug report, spec, or rough idea. It reads the repo, plans the change, writes code, runs checks, and works toward a review-ready pull request. Take a look at vroni.com