RankCaster AI: A GEO Platform for SEO Agencies — November 2, 2026 Release

TL;DR. RankCaster AI’s biggest update adds 7/30/90-day forecasts for which pages AI may cite, a Site Auditor with 600+ AIO/AEO/GEO checks and actionable fix plans, and a complete content workflow from draft to publication. Agencies can connect prompts to search keywords, track AI citations by article, and automate work through MCP and publishing integrations.

Six weeks after the previous release, RankCaster AI has received its largest update since the platform launched. The platform now does more than show where a brand appears in AI answers: it helps predict where to place content, runs a detailed audit with a step-by-step plan to improve AI search visibility, and provides a complete content workflow from idea to publication.

RankCaster AI release by the numbers

450 commits and 340 merged PRs
More than 600 unique auditor rules across 3 categories (AIO / AEO / GEO) and 12 analyzers
4 prompt data export formats
7 Awareness Ladder levels (L0-L6)
Approximately 15 notification types
5 publishing adapters (WordPress, Medium, Wix, Dev.to, Headless API)
7 report languages (English, Spanish, Russian, Hebrew, Arabic, Ukrainian, Georgian)

Citation Opportunities: forecasting brand citations in AI search

RankCaster now predicts which pages AI is likely to cite next, for which queries and on which platforms. This is not analytics about “what already happened”; it is a forecast of “what may happen in the next measurement cycle.” The closest analogy is a weather forecast: not a guarantee, but a probability calculated from real data rather than intuition.

How citation forecasting works

We take a prompt and candidate URLs and run them through a model trained on our own measurement data. It considers hundreds of signals, from domain age and schema.org coverage to the vector similarity between the content and the prompt. The result is a number from 0 to 1: how likely AI is to cite that page for that query.

The Source Authority Model (ASAM) is still in beta, but it is already available. The model is retrained on fresh data every month, and scores are recalculated daily, so forecasts become more precise from one cycle to the next.

What users see

The Opportunities section lists sites where the brand is not present yet, but where AI is likely to look for an answer. Each opportunity includes a citation likelihood score and forecasts for 7, 30, and 90 days.

Why GEO agencies need it

Choose where to place content and whom to approach for publication based on numbers, not guesswork.

ASAM data for administrators

The admin panel includes a full ASAM citation map: which domains AI cites most often, how they are connected, and how scores are distributed across the market. We are keeping it private for now while we consider the implications of making a ready-made market map public for SEO and GEO.

Site Auditor: a one-click GEO audit

Run a site audit and get a ready-to-use report: more than 600 checks, 12 analyzers, three summary scores from 0 to 100 for AIO, AEO, and GEO, plus an overall score. This is not an abstract “SEO score.” It answers a concrete question: “What is stopping this site from appearing in answers from ChatGPT, Perplexity, Gemini, Claude, and AI Overviews?”

AIO: can AI access the site?

Is the site allowed in robots.txt for GPTBot, ClaudeBot, PerplexityBot, and other crawlers? Does it respond to their User-Agents? Is noindex enabled? Are canonicals correct? Does the content exist only after client-side rendering? The audit also checks Core Web Vitals and sitemap status. If AI cannot read the page, nothing else matters.

AEO: what will AI say about the brand?

The audit checks schema.org validity (Organization, Product, LocalBusiness, FAQ, Article), required fields, product identifiers, return and shipping policies, and Open Graph. It also checks the Knowledge Graph: sameAs links to Wikidata, Wikipedia, LinkedIn, and Crunchbase-the signals AI uses to verify that it is looking at a real entity.

GEO: is the page suitable for citation?

The audit checks heading structure, paragraph length and chunkability, lists and tables, E-E-A-T, freshness, and click depth from the homepage. It also scores each page’s helpfulness, clarity, and completeness from 1 to 5, with a short explanation.

What users get

A ready-to-share report for selling or defending a budget. The HTML report includes a “severity × effort” matrix and a “quick wins” section, so priorities are clear at a glance. Export it to PDF in one click and attach it to a client email, send it to legal or marketing for budget approval, or give developers concrete instructions.
The reasoning behind every rule. Each check includes two sections: [why it matters]-a plain-language explanation of its business impact to use with a client or CMO-and [extended fix]-step-by-step instructions for what to change in code, markup, or content for a developer or content manager.
A critical gate. If the homepage has noindex or Googlebot is blocked, the overall score does not get averaged out by middling results; it drops into the red zone. The client sees a high-priority alert instead of simply receiving an “overall score of 72.”
Connect the report to Claude, ChatGPT, Cursor, or any other MCP-compatible agent. The agent can access your reports as its own tool. You can say, “Take the latest audit for site X, go through the critical findings, and make the fixes on the site.” The agent retrieves the recommendations, works through them by priority, and writes the code.

The audit outcome for agencies and clients

Instead of “We looked at the site; you need to redo your schema,” you get a prioritized document with fixes that you can give to the client today, add to your agency’s AI pipeline, and use as the basis for next week’s work.

Content Workflow: from draft to approval and publication

Raw AI text detector

Every generated draft is first checked for signs of unedited AI text: sentences of similar length, stock transitions, and tired phrases. If the text does not pass the threshold, the system regenerates it with a different configuration before sending it to an editor. No “raw” article goes out for approval.

Adapt for…, one article, four channels

With one click, the main piece becomes versions for LinkedIn, email newsletters, X/Twitter, and short-form video (TikTok, Reels, Shorts). All versions are stored beside the original and versioned, so you can always return to an earlier edit.

Approval through a private link

Send a link to a client or external editor. They can open the article, leave comments, approve it, or request changes-without registration, passwords, or workspace invitations. Decisions and discussion history stay on the article card, while anything awaiting a response appears in the Activity Inbox. Nothing gets lost.

An editor built for long-form content

The action bar (publishing, status, prompt coverage) stays pinned at the top and remains visible as you scroll. The Schema.org card shows which JSON-LD will be added to the page, and you can edit it without leaving the editor.

Content Manager: reporting and performance for every article

The header shows four production-wide numbers: total assets, items awaiting approval, articles published this month, and total AI citations over the selected period.

Each article has an AI citation counter. Open an article to see how many times ChatGPT, Perplexity, Gemini, Claude, and AI Overviews cited it; which prompts triggered the citations; which providers cited it; and when it last happened. This is not traffic or search rankings. It is a direct count of “how many times AI used your article to answer someone.”

That answers the client’s big question: “Are you actually doing anything with our money?” An article with 847 citations in a month can defend its own budget. An article with zero citations honestly shows that the hypothesis did not work-and points to the next one to test.

The old world of SEO measured work by Google rankings. The new world has a direct counter: how many times AI decided your content was the best answer. That counter sits beside every asset.

Blog API and publishing to external platforms

A built-in blog for every project

Every client project automatically gets its own public space for publishing articles created on the platform. Manage several brands in parallel, each with its own project, blog, content, and domain.

Your own domain with one CNAME

Add one record to the client’s DNS and the blog starts resolving on their domain (blog.client.com). Setup instructions appear in settings as soon as the report is created-no DevOps tickets and no waiting.

Headless JSON API

If the agency does not want to use RankCaster AI’s public blog, take the content with you: it is served through a JSON API and can be added to any frontend-Next.js, Nuxt, Astro, or anything else. Full documentation is on the report’s Integrations page, with endpoints, schemas, and request examples.

One-click publishing channels

WordPress
Wix
Dev.to
Headless Blog API

Channels connect through their native REST APIs and Application Passwords. Articles publish directly, without copy-paste, with all key SEO data: canonical URL, JSON-LD graph, keywords, and phrases. Integration setup instructions are available in 5 languages.

The publishing menu shows connected channels. Publish an article to dozens of connected sites and blogs in one click, without copy-paste; revise it and publish again. It saves content managers days of work.

Keyword Bridge: from AI prompt to page optimization

Until now, marketers had two separate reports: one showed the queries in which AI mentions a brand; the other showed the keywords people search for. There was no bridge between them. Decisions relied on intuition: “This keyword seems right,” or “This topic seems important to ChatGPT.”

In this release, we connected both layers in one workflow. Each prompt tracked in Prompt Intelligence now generates a list of search queries, their search volume, and ready-to-use recommendations for Content Manager-what to add to a specific page so it can start appearing in AI answers.

What happens automatically

Keywords are extracted from your prompts. As soon as you add a prompt to Prompt Intelligence, the platform pulls out the search queries people actually use. There is no separate keyword research step: the keywords come from what you already monitor.
Search volume is collected by region. RankCaster checks the search volume for each keyword in the selected region.
Keywords are compared with cited competitors. For each keyword, the platform calculates what share of the most-cited competitor pages actually contain it. This metric, AI Share, answers: “Does this keyword work in AI results for this prompt?”
Content Manager gets ready-to-use recommendations. If a keyword has (a) meaningful search volume and (b) appears on most cited pages, Content Manager suggests adding it to your page. It is deterministic, without “AI magic”: a rule you can show and explain to a client.

What changes in practice

Keyword research stops being a separate service. Keywords come from what you already track. You pay for monitoring and get a list of queries to feed into your content strategy at no extra cost.
Recommendations are supported by evidence. Each one is based on two numbers: search volume and share of cited pages. The client can see why you chose that keyword instead of hearing “because we said so.”
One asset serves two channels. The same page work affects both traditional search and LLM visibility. You no longer need separate strategies for Google and AI.
The feedback loop is closed. Every page change is checked against the following week’s measurements. If a keyword did not improve citations, you will see it instead of finding out six months later.

MCP: work with RankCaster AI from Claude Desktop and other clients

We are introducing a new way to work: RankCaster becomes a workspace inside Claude Desktop and any MCP client, including Cursor, Windsurf, and Zed. You no longer need to open the dashboard to review an audit or edit an article.

Connect once, then work through chat. Copy your MCP endpoint from Connected apps, add it to the Claude Desktop config, and confirm access. OAuth grants access only to your workspace and within your plan.

What you can do in chat

Diagnostics

“Show me the critical findings from the latest audit and how to fix each one.”
“Which domains should I publish on to improve AI visibility?”
“Which prompts is the brand losing citations for right now?”

Content, from draft to publication

“Show me the content topics for the latest project and the drafts for the top-priority one.”
“Rewrite the draft in the brand voice, cover all target prompts, and run the quality gates.”
“The article is live at link X-mark it as published.”

Write, review, and publish without switching to the UI.

What this means for agencies

Content editors handle the whole workflow in one Claude window instead of switching between the dashboard, Google Docs, and the CMS.
SEO specialists can ask for a spoken diagnosis-“What broke for client X this week?”-and get an answer with links to specific pages and rules.
Client updates are prepared in a conversation instead of by manually checking tabs.

It is essentially the same step Figma once took when it turned a design into a document people could work on together. Here, what opens up is not a design file, but your entire layer of AI visibility and content. Available on every plan, including Starter.

APR Monitoring: exports and charts

Export in 4 formats. Prompt data can be downloaded as PDF, XLSX, CSV, or JSON. The PDF includes a small APR trend chart beside each row, ready to attach to a client email. XLSX and JSON provide model-level breakdowns, with separate columns for Claude, ChatGPT, Gemini, Google AI Overviews, Perplexity, and DeepSeek. CSV is for those who want to build their own table.

APR is stored separately for each model, so trends no longer jump around. Previously, overall APR was recalculated on the fly with every request: add Perplexity to the plan and the entire historical curve shifts. Now each model’s APR is stored in its own row, and weights are applied when the data is read. Historical trends remain comparable even if a client changes their provider mix midway through a contract.

Prompt card periods: 7 / 30 / 90 days / 1 year. The chart legend is clickable: turn off ChatGPT and view only Claude and Perplexity, without reloading. The All / Anchor toggle separates deep anchor measurements (10 times a day) from one-off daily runs, so you can view a clean trend without daily measurement noise.

PDF reports in your chosen language

PDF export has been redesigned: a branded cover, six sections, and wide tables that automatically switch to landscape orientation.

The report language is independent of the interface language. Ukrainian and Georgian have joined English, Spanish, Russian, Hebrew, and Arabic. Share the finished document with a client without manually fixing tables or translations.

Notifications and sharing achievements

More than 15 event types. These include congratulations for APR growth, alerts about drops, completed measurement cycles, account setup gaps (Persona not added, no monitored prompts, traffic pixel not installed), content approvals and rejections, and weekly cadence reminders. Deduplication per organization ensures the same event is not delivered twice, even if a cron job restarts.

Milestone sharing. When an important metric is reached, a banner appears on the dashboard with LinkedIn and Facebook sharing buttons. Each link opens a polished Open Graph card on a public preview page (/share/milestone). The client sees the result, the press sees the result, and competitors see the result.

AI Visibility Badge. A live SVG badge showing current APR, updated every 24 hours. Embed it on a client’s website or in an email signature with one HTML snippet-like a G2 or Clutch badge, but for AI citations. Every badge on a client site links back to RankCaster, working as both a trust signal and a distribution channel.

Other RankCaster AI updates

Archetype renamed to Persona. The new name is used throughout the interface, API, exports, and translations. Existing links and external integrations keep working: the slug format remains backward-compatible, so your WordPress plugin calling /archetypes/{slug} will continue to work as before.
Awareness Ladder: L6 added. The funnel now runs from “unaware of the problem” (L0) to the post-purchase stage (L6): retaining existing customers and winning competitors’ customers at the moment of greatest dissatisfaction. Seven levels instead of six give the content plan a dedicated place for retention and competitive takeout-areas that used to fall between “purchased” and “next report.”
MCP register_lead-leads from AI chats. This MCP server tool lets Claude, ChatGPT, Cursor, or any connected agent send a lead’s name and email from a conversation to the client’s HubSpot or Brevo. It is enabled with a toggle in MCP Tools and opted into separately for each organization. For agencies, this creates a new lead capture channel: no website form or landing page is needed-the brand’s AI assistant can take the conversation all the way to the CRM.
Stagehand Discovery /aio. Our own Google SERP scraper with an AI Overviews detector. The platform can see whether a client appears in the AI Overviews block without external SERP APIs-no third-party provider limits or separate query charges.

What changes for SEO and GEO agencies using RankCaster AI

Agency workflow before and after the update

Before this release, agency work in AI search relied on heuristics: a hypothesis, a manual report, then an attempt a month later to figure out whether it worked. Now every step is backed by a number you can show the client.

Before	After
“Where should we publish to get the client into ChatGPT?”-a guess based on experience	p_citation + an ASAM-ranked list of domains, with 7/30/90-day forecasts
Keyword research is a separate step and a separate client charge	Keywords are extracted from prompts you already monitor; search volume and AI Share are added automatically
Manual technical AI audit: 4-8 hours per domain	One click → 600+ AIO/AEO/GEO checks and a ready-to-share PDF with a “severity × effort” matrix
“Redo the schema” recommendations with no detail or owner	Every rule includes why_it_matters for the client and an extended_fix for the developer
To get an agent to fix a site, you have to write out tasks manually	Claude Desktop / Cursor / Windsurf connects to RankCaster over MCP, retrieves the audit, and works through findings
Manually reworking each article for LinkedIn, a newsletter, X, and video	Adapt for… button + version history for every variant
Client approval means registering the client in the system	Private link; approve or request changes without an account
“Are you actually doing anything with our money?”-you have to defend the work with words	An AI citation counter beside every article: how many times ChatGPT, Perplexity, Gemini, Claude, and AI Overviews cited it
Monthly client report takes half a day to assemble manually	Content Manager list view is a ready-made report for the selected period
APR trend breaks when the provider mix changes	Per-model storage keeps the historical curve stable when plans change
Publishing to WordPress / Wix / Dev.to means copying and pasting into the client’s CMS	One click to a connected platform; canonical, JSON-LD, and keywords go into the right fields automatically
Lead capture from AI chats is impossible	MCP register_lead → HubSpot / Brevo straight from a brand AI assistant conversation
PDF export breaks Cyrillic and cuts off wide tables	4 formats, branded cover, landscape orientation, and 7 languages including Ukrainian and Georgian
The client sees agency work only once a month in a report	AI Visibility Badge on the client’s site updates every 24 hours-the work is visible every day

A short client pitch

You stop selling “SEO hours” and start selling a measurable result: how many times AI cited the content you created. This is not a future metric-it is a number you can show in your very next call, beside every article you delivered.

A short summary for the internal team

The content editor works in one Claude Desktop window. The SEO specialist gets a spoken diagnosis and fixes the site through an MCP agent. The account manager does not build a report; they open Content Manager and copy four numbers. The time saved goes toward new clients instead of reporting.

Frequently asked questions about GEO platforms and AI search

What is the best GEO platform for SEO agencies and GEO agencies?

For SEO agencies, a GEO platform should connect visibility measurement to action: find citations and publishing opportunities, audit technical and content issues, produce recommendations, and track results. RankCaster AI brings these GEO tools into one workflow: ASAM forecasts likely citations, Site Auditor checks AIO/AEO/GEO, and Content Manager follows the workflow from article ideas through citation tracking.

How can you get your content cited by AI?

Make the page accessible to AI crawlers, and check robots.txt, indexing, canonical tags, and rendering. Improve structured data, the completeness and clarity of the material, freshness, and structure. Then track prompts and competitors’ cited pages, use the content recommendations, and check the changes in following measurement cycles. Site Auditor and Keyword Bridge in RankCaster AI support these steps.

How do you measure revenue from AI search traffic?

To connect AI search traffic to revenue, compare visits from AI sources and conversions in your web analytics and CRM with lead and sales data. RankCaster AI measures brand visibility and citations, as well as traffic from ChatGPT, Claude, and other AI assistants.

Which platform can track AI traffic?

The simplest option is Google Analytics 4: create a traffic source segment using referral domains such as chatgpt.com, perplexity.ai, gemini.google.com, claude.ai, and copilot.microsoft.com. This gives you a basic view of which AI assistant sent visitors, but it does not connect visits to a specific prompt and misses assistant mobile apps, which often lose the referrer.

RankCaster AI covers both layers. Its embeddable pixel (one line of JavaScript) detects every visit from AI assistants-including mobile apps-and shows the source, landing page, and trend in a dedicated AI Traffic Dashboard. On top of that, the platform tracks citations in ChatGPT, Perplexity, Gemini, Claude, and Google AI Overviews. For every article, you can see how many times AI cited it, which prompts triggered the citations, and when it last happened. Together, this gives you the full funnel: AI mentioned the brand → cited a page → a person clicked through to the site.

How can agencies increase AI search visibility for clients?

An agency can start with Site Auditor, remove critical technical barriers, improve pages’ suitability for citation, choose sites using ASAM forecasts, apply Keyword Bridge recommendations, publish through integrations, and measure subsequent citations and APR. This cycle helps SEO agencies justify priorities with data and show clients changes in brand visibility in AI search.

See how AI-ready your site is.