What's new in Trakkr

Actions used to be a long list you sorted yourself. It now opens as a desk: a short written brief on what changed since your last report, a picked plan for the week, the jobs Trakkr is still watching, and the results that have come back. Every job reads the same way, in the same order: what we noticed, what we saw, what to do, and what it should earn. Every number in the evidence is a link to the thing it came from. When you finish a job, Trakkr starts watching that page and reports back later with earned, no change, or could not measure. It will not claim a win it cannot show you.

• Six new kinds of work built from data Trakkr already holds, including cited pages that now 404, review sites shaping your answers, and forum threads where your competitors are and you are not
• Anything a machine can check, like llms.txt, schema, robots rules or a restored page, now verifies itself overnight and closes with a Verified by Trakkr note
• A nightly tidy merges duplicates, drops work whose evidence has faded, and re-states the reason for anything that still holds
• The old two-pane list is still there as the full backlog, with every filter, shortcut and saved link intact

Accuracy has lost its tabs and its setup gate. It is one page now: the verdict first, then what to fix first, then the full ledger of claims AI makes about your business, each one carrying the transcript it came from. Claim titles are checked against the model words behind them before you ever see them, so a headline can no longer borrow a price or a date from the claim next to it. Re-checking is one button. Any run can be exported as a CSV of every claim, or printed as a white-label record for someone who does not use Trakkr.

Optimize used to open with a score and a stack of panels. It now opens with a sentence: what is wrong, on how many pages, and what to do first. Under that sits a short queue of ranked work, each item carrying the measured evidence behind it. The Issues tab is gone. Its full check list now sits at the bottom of Findings, using the same words as the queue, so one problem reads as one problem instead of showing up twice under two names.

• Every count states its base, so you read 74 pages of the 100 we crawled rather than a bare number
• Each item opens to the evidence, the fix, the case where it is a false alarm, and a write-up you can copy straight into a ticket
• Anything we have not measured says so and links to what would measure it, instead of showing a zero
• First run, scanning and stopped scans now use the same page and the same sections as your results

Prompts now have a second view. Give each prompt a topic and the Prompts tab switches from a list to an ownership board: every topic with your share of voice, who leads it, and whether you own it, are contesting it, or are slipping. Trakkr drafts a starter set of topics from the prompts you already track, and you either nod them through or change them. New brands arrive organised, because the topics picked during onboarding carry straight through.

Integrations were spread across several pages that each behaved slightly differently. They now sit in one settings-style hub grouped by category, with what you have connected shown first. Every card says the same things in the same order: whether it is working, when it last synced, and what to do if it is not. WordPress sites show what Trakkr can actually do on that specific site rather than a generic list, and your API key has one home instead of three.

CSV export now comes in three shapes: a summary time series for you and your competitors, a detailed row per model, prompt and day, and the raw top ten for every day and model. Each takes a date range, a model filter and a tag filter, and you can drop columns you do not need. A preview shows the row count and a sample first, so you find out you picked the wrong range before you wait for the file. Past exports stay in a history list.

Setting up AI Pages on Cloudflare could fail quietly, most often because the domain was not proxied through Cloudflare, so the worker never ran at all. The check now tests your site the way a real AI crawler would and reports what it found, and the install guide covers the cases that trip people up.

The data explorer used to give you the number. It now gives you the arithmetic: how each model scored a prompt, how those add up, and how the total moved. Source profiles in Citations were rebuilt at the same time, so following a score down to the pages behind it is one path now instead of three.

We run the same tracking on trakkr.ai that we sell, so we published what it sees: which AI crawlers fetch our pages, how often, which pages they take, how healthy those pages are, and where any of it turns into a citation. It is the same workspace you use inside Trakkr, pointed at us, read only and with no sign-up. If you want to know what this data looks like before you connect your own site, start there.

Perception used to carry a one-word tone label that was worked out from scores rather than from anything a model said. It is retired. Descriptions are now built from attributes with the quote behind each one, so you can see the sentence that produced the read. Agent and Actions were switched onto the same grounded description, which means the work Trakkr suggests is argued from quotes too.

Prompts used to be rows you expanded. Each one now has a full page: the score and its trend chart, a breakdown by model, the sources AI leans on to answer it, the different phrasings real people use, and its Search Console numbers in the side rail. Pressing j and k walks the same filtered order you left the table in, so it behaves like an expanded row rather than a dead end.

trakkr.ai/answers is a growing directory covering how citations work, what moves a model recommendation, and how to measure any of it. Answers are first-party and a person reads every one before it publishes. Free, no account.

A growing number of crawlers pretend to be AI agents and quietly probe sites for exposed environment files, config dumps and key stores. The free scanner at trakkr.ai/data/security probes the same paths on your own domain, read only, and tells you which of them are actually readable from the open internet. It takes a domain and about a minute.

Agencies can now put a client report on a schedule. Pick the client, the cadence and who receives it, and Trakkr sends it: your branding, a live white-label link and the PDF attached. You can hold every send for approval first, so nothing reaches a client until someone on your team has read it. Clients never see Trakkr.

The agency home page is now a single table where every row is a launchpad: how that client is doing, the one thing to do next for them, and a way into their workspace without losing your place. Above it sits a pulse across the whole portfolio and a quiet digest of what moved since you last looked. One control shifts the table between overview, compare and needs attention, so the page never splits into tabs.

Trakkr connects to Search Console the same way it connects to Analytics. Once linked, Google clicks and impressions sit alongside what AI says about you, and they feed the work Trakkr suggests. The pages Google already ranks but AI never cites are usually the best pages to fix first, and that comparison is only possible once both sides are connected.

The rank tracker reads a site, works out the buying questions people genuinely ask AI in that category, asks a model each one cold with no help, and shows where the brand lands or that it does not come up at all. The raw model answer is kept alongside as evidence. Free, no account, and it works on your competitors too.

Trakkr can hold one connected OpenAI ad account per brand. The key is checked with a single read-only call, encrypted at rest, and never shown back to you afterwards. Alongside it we shipped two free things: a generator that finds the buying moments where a brand is losing and writes a paste-ready context hint pack for each, and an open-source MCP server for the OpenAI Advertiser API with 27 tools, read-only first-run defaults and an MIT licence, so you can drive it from Claude, Cursor or Codex.

Perception marked a model as not recognising a brand whenever it hedged. That caught models describing the real business perfectly well while disagreeing with a stale one-line description we had on file. Those now read as recognised, and the models are told up front that the description may be wrong, so they correct it rather than refuse to answer.

trakkr.ai/bias puts the same questions to every major model many times over, with web search switched off, and plots the answers with error bars so you can see how much each model wobbles between runs. Browse by model, by question, or put two models side by side. The full method is published, including what we did not measure.

The Crawlers page has a Map tab. Your site becomes a path tree you can colour by whatever you care about: readiness, traffic, access, citations, freshness or token weight. Collapse it for a bird's-eye view of your sections, switch to journeys to follow the routes crawlers actually take, or to cadence to see how often they come back. Two worklists fall straight out of it: pages crawlers reach but never cite, and pages they never reach at all.

trakkr.ai/best is a public record of what AI recommends. We ask the questions buyers ask across a growing set of sectors, publish the answers in full rather than a summary, profile the products that come up, and name the ones models quietly leave out. Useful whether or not you appear in it, because it shows how a category gets ranked.

Accuracy asks ChatGPT, Claude, Gemini and Perplexity what they tell buyers about your company, then grades every factual claim they make. There is nothing to set up: the facts come from what the assistants actually say. Trakkr keeps the evidence, tracks history, supports re-checks, and helps you turn weak or outdated answers into next steps.

• Covers leadership, ownership, founding, location, pricing, products, competitors and reputation
• Store evidence and check history so teams can see what changed over time
• Hand findings to Agent, action plans, or a printable record when a fact needs attention

Agencies can build sharper, more credible pitch reports from the same visibility signals they use inside Trakkr. Reports can include measured gaps, competitive context, print-ready views, lead gates, and open alerts, with white-label behavior respected throughout the flow.

Trakkr now has a clearer setup path for teams that want their AI assistants to work with live AI visibility data. The new MCP setup flow uses scoped connection tokens, hosted connector support, OAuth consent screens, and client-specific setup guidance so your tools can read the right brand intelligence without handing around raw API keys.

• Guided setup for popular AI clients and developer tools
• Scoped MCP connection tokens for safer access
• Partner OAuth support for read-only brand and AI search intelligence

Weekly reports now rotate through stronger story angles instead of repeating the same summary every week. They can bring forward competitor movement, blind spots, fresh opportunities, and stronger trial context, with safer fallback copy when AI synthesis is unavailable.

Prompt tracking is now easier to act on. Health review, movement pulse, weak-prompt suggestions, new prompt recommendations, and competitor-overtake diagnostics help you see which prompts deserve attention and what to do next.

The Citations workspace has moved beyond watching where your brand appears. You can now work through richer source profiles, query intelligence, outreach opportunities, video citations, owned-channel linking, accuracy and sentiment filters, and export paths that make citation work easier to share with the team.

• Richer source and query context for understanding why citations happen
• Outreach and owned-channel paths when a source deserves follow-up
• More ways to filter, export, and hand off citation work

Agent has grown into a full workspace for visibility work. It supports threads, memory, quick and deep modes, structured streaming answers, tool-backed analysis, confirmable suggestions, CSV exports, and better evidence guardrails. It is less brittle, more grounded, and much better at staying with the shape of the question you actually asked.

• Ask about visibility, citations, crawlers, prompts, actions, and site opportunities in one place
• See tool activity and pinned context so answers are easier to trust
• Review and confirm suggested next steps instead of copying instructions into another workflow

The start flow now does a better job showing what Trakkr is learning about a brand before the first full report lands. It can surface site-specific observations, crawler readiness signals, report progress, model chips, stalled-scan recovery, and checkout recovery in a calmer flow.

Crawler data now feeds a real action loop. Trakkr can review how AI crawlers are reaching your site, inspect the pages they see, and group recommendations around leaks, conversions, and expansion opportunities. Instead of staring at logs, you get a clearer view of what deserves attention next.

• Agentic crawler analysis grounded in what bots actually receive
• Recommendations grouped by opportunity type
• A closer link between crawler behavior, page readiness, and Actions

Teams now have a clearer place to check component health, incident states, maintenance windows, and freshness safeguards. The status experience is built from sanitized snapshots, so operational updates can be shared without exposing internal detail.

The crawler workspace now gives teams more ways to understand how AI systems interact with their site. Page-level drawers, bot and citation context, source and verified-session filters, date ranges, trend views, CSV exports, and clearer connection states make crawler analysis much easier to use in real work.

The Create Agent watches the signals you care about and turns strong opportunities into draft work. You can run it manually, let it work on a schedule, choose what it watches, set quality gates, review the drafts it keeps, and allow high-confidence drafts to publish automatically when your settings permit it.

• Automation settings for pace, sources, quality gates, and publishing behavior
• Review queue for Agent-written drafts
• Citation-likelihood scoring before drafts move forward

Optimize can now draft changes from audit issues, show a reviewable diff, apply approved updates through connected destinations, and check the result afterward. It is designed to keep humans in the loop while removing the slow copy-paste handoff between diagnosis and implementation.

• Review proposed changes before anything is applied
• Apply approved updates through connected CMS or GitHub destinations where supported
• Track applied, verified, declined, and rolled-back proposal states
• Handle multi-page recommendations without opening each page from scratch

The Optimize workspace has been redesigned around the way teams actually review site work. The new layout makes it easier to move from the big picture to individual pages, compare what changed over time, expand issues inline, and spot the recommendations with the highest practical value.

Site connections are much more capable. Teams can connect publishing destinations, configure how Trakkr sends drafts or updates, and use GitHub with safer framework-aware route mapping. For code sites, Trakkr can explain why it picked a file, show confidence and evidence, support manual overrides, and open reviewable pull requests instead of guessing.

• Publishing flows for connected site destinations
• Framework-aware GitHub route mapping and PR targeting
• Manual review paths when Trakkr does not have enough confidence to act automatically

Create is now a full workspace for finding, shaping, and measuring AI visibility content. It brings together citation gaps, prompt opportunities, Reddit demand, crawler signals, GA4 context, manual ideas, campaigns, draft generation, and cited-article tracking so content work starts from evidence instead of a blank page.

• Opportunity lists powered by multiple visibility signals
• Campaigns, manual ideas, quick generation, and bulk actions
• Citation measurement so teams can see which live articles are starting to earn mentions
• Evidence, coaching, exports, and settings for brand voice, knowledge, templates, and site context

Actions and operational work can now move into external tools with more structure. Teams can set destinations, routes, queues, and status tracking for tools like Linear, GitHub, Slack, Trello, Notion, Asana, and Jira, depending on the connection available to the workspace.

For pages that need a nudge after important updates, Trakkr can submit eligible URLs through IndexNow and show recent submission status with clear success or error feedback. It gives teams a cleaner handoff after publishing important AI visibility changes.

Dashboard, Citations, Competitors, Diagnose, and Prompts now share more consistent location and scope behavior. For brands tracking multiple markets, this makes it easier to keep regional questions and regional reporting lined up across the app.

Agent gained durable multi-step plans for work that should not happen in a single chat response. Plans can show progress, pause for approval, support retry or dismiss actions, and check back after work completes, giving teams a clearer way to collaborate with Agent on larger tasks.

Google-Agent is fundamentally different from traditional crawlers. Instead of reading pages, it fills forms, navigates, and clicks - representing AI agents acting on behalf of users. Trakkr tracks it as its own entity, distinct from Gemini (training) and Google (search). If you use Cloudflare Workers or Netlify Edge templates for crawler tracking, regenerate your snippet to start capturing Google-Agent visits.

• New display entity: Google Agent appears separately from Gemini and Google in all charts
• New bot category: "AI Agent" - the first of its kind, distinct from crawlers and search bots
• Prism serves optimized content to Google-Agent by default
• Updated setup templates - regenerate your edge function to capture Google-Agent

We've renamed Prism to AI Pages. The feature works exactly the same - intercepting AI crawler requests and serving them optimized versions of your pages with structured data, summaries, and FAQs. The new name better describes what the feature does: every page on your site gets an AI-optimized version, served to crawlers automatically.

A new pentagon radar overlay in the Perception tab lets you visually compare your brand against any competitor across trust, expertise, value, innovation, and market presence. The chart animates in with framer-motion and updates instantly when you switch competitors - making it easy to spot exactly where you lead and where you trail.

Citations now include a source type dropdown that breaks your mentions into owned properties, earned media, social, reviews, institutional, competition, and PR wire. Each type shows a live count so you can see your citation mix at a glance, then drill into the types that matter most. The old source type bar has been replaced with a cleaner, more compact filter.

We've added seven new endpoint families to the public REST API. You can now pull perception dashboards and trigger analyses, manage narratives (full CRUD), run AI visibility diagnoses with entitlement-based limits, fetch and refresh content ideas, list and generate reports, query crawler activity, and retrieve rankings by prompt across markets. Every new endpoint includes rate limiting, structured response schemas, and dedicated documentation.

• Perception: dashboard view, metrics view, trigger analysis
• Narratives: list, create, update, delete
• Diagnose: run analysis, retrieve results (with monthly limits)
• Content Ideas: list current ideas, trigger refresh
• Reports: list reports, generate new report
• Crawler: dashboard and recent crawl views
• Rankings: overall scores and per-prompt breakdown by market
• Brands: add/remove markets, update aliases

Reports can now be scoped to individual markets. Add multiple regions to a brand, and each report run will geo-target its queries - Perplexity and OpenAI get localized search results, Google AI Overviews use the correct DataForSEO location code, and cache keys are scoped per region so results never bleed across markets. Great for global brands that need to understand AI visibility in each country independently.

• Per-brand market management (add, remove, set primary)
• Geo-targeted queries for OpenAI, Perplexity, and Google AI Overviews
• Per-prompt location overrides for mixed-region tracking
• Location-scoped caching so regional results stay separate

We've built a full content creation environment right into Trakkr. Everything you need to produce AI-optimized content without leaving the platform.

• Knowledge base for storing brand information and messaging
• Reusable content templates you can customize
• Brand voice and tone configuration
• AI audience profiles that write in different styles

This is a major upgrade to our original sentiment analysis system. We've rebuilt it from the ground up to give you a clearer picture of how AI models perceive your brand. You'll see real-time tracking across all major models, historical trends you can explore with interactive charts, and direct comparisons with your competitors.

Set up automations that run when things change in your AI visibility. The visual builder makes it easy to create workflows with drag-and-drop, connect them to Slack, email, or webhooks, and add conditional logic to fine-tune when they fire.

• Pre-built templates for common scenarios
• Conditional logic and filtering
• Slack, email, and webhook integrations

Meet your new AI visibility assistant. Ask questions about your brand data in plain English, get proactive insights and recommendations surfaced automatically, and take quick actions through action cards. The Agent remembers your conversation history and context, so you can have natural back-and-forth discussions about your data.

We've massively expanded what you can see in citations. Every source now has an authority score so you know which mentions matter most. Queries are automatically clustered into patterns so you can spot trends, and we surface outreach opportunities where you're being mentioned.

We've merged our Optimised Sites and Technical Audit features into one comprehensive solution. Your site now gets a thorough weekly audit that covers both technical health and content optimization, with your score tracked over time and clear recommendations you can action immediately.

The competitors tab has been completely rebuilt. You can now create custom breakdowns to see exactly what you need, slice the data by prompt or by model, and get detailed positioning information for each competitor.

• Custom breakdown configurations
• Analysis by prompt and by model
• Side-by-side competitor comparisons

We've significantly upgraded our site analytics. You can now see live visitors on your site in real-time, plus detailed tracking of AI crawlers - which bots are visiting, how often, and what they're looking at. Great for understanding how AI models are learning about your brand.

Big expansion to how teams work in Trakkr. We've added much more granular access controls, plus two major new capabilities:

• Client workspaces - add clients to specific brands with their own permissions
• Brand groups - organize brands under parent entities
• Team activity feeds for better collaboration

Not all prompts are created equal. Demand scoring helps you understand which prompts matter most based on how often they're being asked and their potential business impact. Focus your optimization efforts where they'll make the biggest difference.

We've made it much easier to discover new prompts worth tracking. The Prompt Ideas section on the Prompts page now surfaces more relevant suggestions based on your industry, with better navigation and filtering to help you find exactly what you're looking for.

Need to share AI visibility insights with leadership? Executive Reports generates polished, presentation-ready reports automatically. Customize the templates to match your needs, and share dashboard views directly with stakeholders.

We've launched a dedicated learning center to help you get the most out of AI visibility.

• In-depth guides and tutorials
• Strategy playbooks for different industries
• Case studies showing what's working for other brands

You can now connect your WordPress site directly to Trakkr. Once connected, the Agent can suggest changes and even apply them for you - making it faster than ever to act on optimization recommendations. Webflow support is coming soon.

The new Activity section gives you a quick overview of everything happening in your account - scans completing, scores changing, team members taking actions. It's all there in a simple chronological feed you can filter to find what you need.

Our pitch generation has gotten a major upgrade. The quality is noticeably better, with smarter matching to journalists and outlets who are actually likely to cover your story. The whole workflow is more streamlined too.

You'll notice Trakkr looks different. We've moved to a new bento-style grid system that lets us show more information in less space, refined the typography and spacing throughout, and rebuilt many of the dashboard components. The mobile experience is much improved too.

We've done a lot of work under the hood to make everything faster. Pages load quicker, data fetching is more efficient, and interactions feel snappier. You shouldn't have to wait around for Trakkr anymore.

For developers integrating with Trakkr, we've shipped a completely new REST API. It's faster, better documented, and includes webhook support for real-time updates. Rate limits have been improved across all plans too.

We've rebuilt the audit experience from scratch with a new step-by-step wizard that makes it easier to work through recommendations.

• Actionable recommendations with priority scores
• Export audit results as PDF
• Track implementation progress over time

Design content templates optimized for AI citations. The visual editor lets you build section blocks, and we've included built-in templates for common content types. Each template gets an optimization score so you know how well it's structured for LLM citation.

Discover publications covering your space and see which outlets cite competitors but not you.

• Build targeted media lists
• Track outreach progress

We've added a suite of features designed specifically for agencies managing multiple client brands.

• Brand comparison dashboards
• Client demo mode
• PDF export for reports
• White-label slide decks

An interactive knowledge graph that shows how AI models connect your brand to topics and competitors. You can track how these associations change over time.

Citation tracking now includes domain authority scores for each source, plus advanced filtering by source type and date. You can also export everything to CSV for reporting.