Bot View
Bot View shows the readable text in a public page response. Compare it with the text available after JavaScript runs to find information that some crawlers may miss. It does not change your website, CMS, or the experience of human visitors.
Open Bot View
Section titled “Open Bot View”Sign in and open Bot View. You do not need a paid subscription, a completed brand setup, a traffic integration, or a scored GEO Audit to inspect a public page.
You can also select View bot content on a page in GEO Audit > Pages, including a discovered page that has never received a scored audit. The same viewer opens from Bot View links in Traffic > LLM Analytics.
Opening a page reads saved evidence only. It never starts a website request or a browser comparison. If no preview exists, select Check page. A missing preview does not mean the page is empty or blocked.
Read a page, step by step
Section titled “Read a page, step by step”- Enter the complete public page URL and select Open page.
- Choose a request profile. ReachLLM is the general inspection profile.
- Select Check page to capture the initial response, or read the saved preview and its capture time.
- Choose Compare with browser to capture the initial HTML and the page after JavaScript from the same clean browser visit.
- Open Initial page content, After JavaScript, or Differences to compare the evidence.
- Review the access rules by purpose and the inspection limitations before deciding what to fix.
- After a website change, select Check again, or run another browser comparison, to collect new evidence.
The readable preview includes page text. Headings, links and structured data are available in the capture details. Text already present in HTML, including a closed accordion, is not automatically missing just because a human needs to click to display it.
Browser comparisons use the ReachLLM request option and a normal browser user agent. The comparison's initial response and final text come from the same navigation. A previous bot simulation is not used as its baseline. Standalone initial previews and paired browser comparisons are saved separately, so their times can differ.
What the evidence means
Section titled “What the evidence means”| Evidence | What it establishes |
|---|---|
| Initial page content | Text extracted from the response received by ReachLLM without running page JavaScript. |
| Saved technical crawl | Text retained during the free GEO page crawl, with its original capture date. Older crawl summaries may have no saved text preview. |
| Browser comparison | Differences between the initial response and the page after JavaScript during one ReachLLM browser visit. |
| Bot profile simulation | The response to a ReachLLM request using the selected user-agent name. It is not a visit from that provider. |
| LLM Analytics request | A recorded request with a path, time, claimed crawler identity, and available status or timing. These logs do not contain the response body. |
For example, a difference may show: "The pricing explanation appears after JavaScript and is missing from the initial response." That text may be missed by crawlers that do not render the page. It does not prove that ChatGPT cannot read it or that a particular provider omitted it from an answer.
Page warnings and exact fixes
Section titled “Page warnings and exact fixes”Every saved complete browser comparison includes a page diagnosis. A warning appears when 25% or more of the compared text is absent from the initial response, provided the page has at least 100 normalized comparison words. This is ReachLLM's product warning threshold, not an AI-provider standard. GEO Audit's page inventory shows the saved warning beside that exact page. Open View bot content to see the dated capture, missing passages, and recommended fix steps. The warning is informational and does not change the technical score or its issue count.
The initial text coverage percentage measures text overlap in the same browser visit. Words count as present when they participate in a matching consecutive three-word sequence, or an existing complete readable line of fewer than three words. Case, Unicode normalization, and punctuation are normalized. Existing reordered paragraphs and repeated navigation do not become missing just because their position changes. The percentage uses the full comparison text, not the shortened list of example passages. It does not measure content importance or the percentage a provider can read.
Smaller observed differences still show the exact passages for review. Short pages show passages without a percentage warning. Incomplete captures, shortened text, access challenges, errors, unavailable legacy evidence, and redirects outside the brand's scope do not receive a percentage. A shortened HTML source display alone does not make otherwise complete text evidence incomplete.
A saved initial crawl without a browser comparison says Browser comparison needed. The product cannot know how much JavaScript adds from an initial word count alone. Opening the inventory reads saved diagnostic metadata only. It does not launch browser comparisons across the site. After making a change, run another comparison for the exact URL to verify the important passages now appear in initial page content. Capture dates and the seven-day evidence lifetime still apply.
An access challenge, an HTTP error, a non-HTML response, and a failed inspection stay distinct from a readable page. Unknown access evidence does not become an allowed result. Automated profiles and browser comparisons check robots policy before requesting page content. If policy blocks the check or cannot be established, the viewer explains that no page-content fetch was made. This is not evidence that the page is empty or that the server returned an error.
Different bots have different purposes
Section titled “Different bots have different purposes”The profile picker separates search discovery, user-requested retrieval, training, and general inspection. OpenAI search, ChatGPT user requests, and OpenAI training use different profiles. Claude and Perplexity also have separate search and user-requested profiles, with Claude training shown separately.
The robots panel shows matching allow or disallow rules for the page path. A training restriction expresses a different preference from a search restriction. Blocking a training bot is not automatically a search visibility problem. Robots rules describe policy and do not prove that the server will permit or deny a request.
User-requested retrieval profiles are explicitly initiated manual simulations. Their robots rules are shown separately from the response; they do not establish whether a provider would honor that policy in the same way.
Google-Extended is shown only as a policy control. It is not an HTTP request profile. A Googlebot simulation does not establish a visit from Google AI Mode or AI Overviews.
All simulated requests originate from ReachLLM. Provider IP addresses, geography, cookies, private extraction methods, and other behavior can produce different results. No preview proves training, model retention, citation, ranking, purchase, or answer inclusion.
Browser comparison limits
Section titled “Browser comparison limits”The browser starts without your account cookies or sign-in. It does not click consent banners, scroll, or interact with the page. It blocks images, media, fonts, frames, workers, and non-GET requests, and applies time, request, and transfer limits.
Delayed content, third-party content, cookie-dependent text, and pages that rely on blocked requests may be incomplete. Read the limitations and any blocked-request or timeout notice before treating a difference as a JavaScript-only content problem.
If an important script, stylesheet, or data request fails, or a capture limit prevents completion, the report is marked incomplete. It can show the failed request and reason, but does not present a completed difference result. An unavailable comparison never means that no differences exist. Intentionally skipped images, media, fonts, and frames are listed as inspection limits rather than evidence of missing text.
Free access and saved evidence
Section titled “Free access and saved evidence”Each signed-in user receives 10 initial page checks and 3 browser comparisons per day. The allowance resets at midnight UTC. A browser comparison has its own allowance and does not consume a scored audit page or execution credits. These inspections do not call an AI model.
Opening saved evidence is free and does not use the inspection allowance. An inspection can reuse the same saved capture for 24 hours when a fresh check was not requested. A requested fresh check uses the relevant daily allowance. Browser comparisons also share a service capacity limit, so a new comparison can be temporarily unavailable even when you have personal allowance left.
The latest saved captures remain available for 7 days. A new capture replaces the previous capture for the same page and request mode; Bot View is not a permanent version history. Previews are private to the signed-in user, or to authorized members of the brand when opened in that brand's scope. Saved GEO Audit pages stay within the current brand and workspace, including older audits and saved links. If your current brand has no saved audit, GEO Audit shows an empty state instead of substituting another workspace's results. You can still use Inspect URL for a free Bot View. A matching website in a different workspace does not grant access. A brand view does not expose another user's personal previews.
Open content from traffic
Section titled “Open content from traffic”In Traffic > LLM Analytics, select Bot View beside a top page or recent request. A recent-request link preserves the logged time and bot label.
The viewer identifies the historical request separately from the preview's capture time. It cannot reconstruct the body sent during that visit. A successful request does not establish what the provider retained, and a simulation returning an error does not prove that the real provider was blocked.
Ask the ReachLLM agent
Section titled “Ask the ReachLLM agent”When the Agent is available for your account, you can ask:
- "Explain the saved Bot View for our pricing page."
- "Compare this page with a browser and show what appears after JavaScript."
- "Claude requested this page yesterday. What do the logs actually show?"
- "Is this robots rule blocking search or only training?"
The Agent can read saved previews and connected request logs, request a bounded inspection when you ask, and link to the same evidence. It uses saved evidence first and does not run another check just to explain an existing preview. Its ordinary conversation entitlement still applies even though the underlying Bot View inspection is free.
Ask "What should I fix on this page?" for the saved diagnosis, exact missing passages, measurement limits, and a verification plan. If you explicitly ask it to fix the page, the Agent can use the existing connected Website, CMS, or GitHub tools where they support that change and your access permits it. It should use observed public facts and the current page source, without inventing prices, claims, or features. If the connection cannot change the site's rendering configuration, it prepares precise developer instructions instead of claiming the website was fixed. Publishing retains each existing tool's permission and confirmation rules. A proposed edit, published change, and verified improvement are different states.
For a question about several pages, the Agent can read saved diagnostic summaries for up to 50 known brand URLs from your GEO inventory at a time. It states the requested-page coverage and keeps pages with no current comparison unknown. This read does not discover URLs, request website content, or run browser comparisons.
Turn a finding into a useful fix
Section titled “Turn a finding into a useful fix”If important public information appears only after JavaScript, make it available in your site's initial HTML through supported static or server rendering. If an unintended rule blocks a search crawler, review that specific rule. If an inspection receives an access challenge, review the website's access controls while preserving the intended protection.
Recheck after the change and compare the observed passage or rule. Bot View does not publish website changes or provide managed bot rendering.
What rendering means
Section titled “What rendering means”Rendering turns page code and data into the content a visitor can read. Browser rendering runs JavaScript after the initial response. Static or server rendering includes the useful content in that initial HTML, so a parser can read it without executing page scripts. For example, a blog's product explanation can be visible in a browser but absent from its initial response if JavaScript inserts it later. Bot View identifies that gap and shows the actual passages to address.
Managed bot rendering would operate in the website's delivery layer and serve a rendered version to recognized automated visitors. That requires installation, safe cache and identity rules, content parity, and operational monitoring. It remains deferred. You can already diagnose a problem, prepare supported source changes through the Agent, and verify the published result with Bot View.
See GEO Audit, LLM Analytics, and the ReachLLM Agent for the related workflows.