Changelog

​

Link building

• Import rows from a spreadsheet: Open a link-building activity and choose Import rows to append rows from an .xlsx or .csv. Columns are matched to your activity’s own headers (order, case and spacing don’t matter), and a misspelled header is flagged before anything is imported. Rows are always appended, never replaced.

​

Custom dashboards

• Three new AI Visibility widgets: Brand presence by prompt, Top cited domains and Competitors mentioned can now be placed on any custom dashboard, each with a row-count option. They render exactly as they do on the AI Visibility report.
• Resize corner visible in dark mode: The widget resize handle was nearly invisible on a dark card. It is now clearly visible and larger.

​

Agent browser

• Browser panel on the task: While an agent is driving a browser, a Browser panel sits beside the task discussion with a live view and the page it is on. It folds away when the agent is not browsing.
• Read-only by default, take control when needed: Watching never interferes with the agent. If you need to step in (a login prompt, a dialog the agent can’t get past), take control for a ten-minute window, then release the browser back to the agent and its follow-up run starts immediately. Taking control is recorded in the activity log.
• Agents can ask for a person: A stuck agent can now hand the browser to you and stop, rather than waiting or guessing. The session stays open long enough for you to act.
• Cookie and consent walls: Agents can dismiss cookie and notification overlays that were blocking forms. They decline rather than accept where a choice is offered, and report whether the overlay actually went away.

​

Help

• Report a bug: Under Help & FAQ in the account menu there is now a Report a bug button. Reports reach the team directly, and you are only told it was sent when it actually arrived.

​

AI Visibility

• Every prompt gets its turn: Projects with more enabled prompts than a single run covers were re-checking the same newest prompts every week and never reaching the rest. Prompts now rotate so every enabled prompt is checked in turn, at the same cost per run.

​

Content editor

• Internal links recognised as full URLs: A link to your own domain written as a full URL now counts as an internal link in the SEO score instead of an external one.
• Image controls: Click an image in a draft to replace or delete it; replacing keeps the alt text. The SEO panel also has a close button on desktop.

​

Agents

• Screenshots as evidence: An agent can attach a screenshot to its task, so a submitted form or a page it acted on is recorded as an image you can open later, not only as a claim in a comment.
• Outages no longer stop a task for good: Runs that never reached the AI model (a provider outage, a restart mid-run) no longer count towards a task’s run limit. Tasks that had been silently stranded resumed on their own.
• Blocked pages are reported, not guessed around: When a page blocks an agent (for example with a CAPTCHA it cannot clear), it waits briefly, then marks the task Blocked with a comment telling you what is in the way, instead of inventing a next step or claiming success.
• Topic Research reads what you already publish: The Topic Research Agent now reads your sitemap and checks every candidate topic against your published pages, so it stops proposing subjects you already cover.

​

Schedules

• Persistent failures pause with a reason: A schedule that fails eight times in a row now stops, with a plain-language reason on its card. Run now clears the pause.
• One bad competitor entry no longer breaks detection: A mistyped competitor domain (for example one without a .com) used to fail the whole gap detector for the project. Unusable entries are now skipped and the field refuses them on save.

​

Agent browser

• Interactive browsing: Agents can open a page, read it, wait for it to load, click, type, choose from dropdowns and press keys in a real browser. Reading is available to every agent; filling and submitting forms is granted to the SEO Outreach and Backlink Building agents.
• Watch live from the run page: A run’s detail page gains an Agent browser panel with the current URL and a Watch live view that follows the agent from page to page.
• Saved sign-ins for agents: Project settings has a new Sign-ins tab. You sign in to a site yourself in a live browser and save it for that domain; agents working tasks in that project browse with that session. Your password is never seen or stored, and only tasks in that project can use the sign-in.
• Sensitive typing is never logged: Anything an agent types into a password or secret field is redacted from the run transcript; only its length is recorded.

​

Tasks and chat

• New comments are answered, not the original brief: An agent woken by a new comment on a task was re-running the task’s original description. It now treats your latest unanswered comment as the live instruction.
• Content editing from chat: You can ask an agent in Chat to edit a specific draft; content tools were previously only available inside a task.
• Transient AI provider failures are retried: Timeouts and temporary provider errors are retried automatically instead of failing the run on the first attempt.

​

Keyword Ranking

• “50+” instead of a dash: A keyword that was checked and not found in the top 50 results now shows 50+. A dash means the keyword has not been checked yet. The two used to look identical.
• Rank checks through DataForSEO: Keyword rank checks now run through DataForSEO, looking through the top 50 results for each tracked keyword.

​

WordPress publishing

• Native blocks: Drafts published to WordPress now arrive as proper blocks: paragraphs, headings, lists, images and more, each editable in the block editor. No more single Classic block and no Convert to blocks step.

​

Tasks

• Files referenced right in the comment: When an agent refers to a file on the task (a spreadsheet it read, a report it produced), the comment shows the file itself with its icon, size, download link and a thumbnail for images, instead of “see the attached file”.
• More attachment types: Task attachments now accept legacy Office files (.doc, .xls, .ppt), macro-enabled and template variants, OpenDocument formats, RTF, TSV, and gz, tar, 7z and rar archives, alongside PDF, zip and modern Office files. CSV files uploaded from Windows are no longer refused as corrupt.

​

Opportunities

• Faster retry for gap detectors: A gap detector that fails is retried within the day rather than waiting for the next weekly run.

​

Team and roles

• Pages follow your role’s permissions: Every page and button now checks the specific permission your role holds rather than whether you are an admin. Members on custom roles such as Project Lead or Manager can now use the features their role grants, and no one is shown a control that would be refused.

​

Agents

• Audits only cover pages your site lists: Technical and on-page audits start from your sitemap (or your homepage’s links when there is none) and never audit a URL the agent made up. A page missing from your sitemap is reported as a finding.
• No invented image URLs: Content-writing agents no longer insert image links they cannot verify. At publish time, an image that definitively does not exist is dropped and you are told, so a broken picture never ships.

​

Content editor

• Pull from URL refuses Visibility’s own links: Pasting a Visibility task link into Pull from URL used to import the sign-in page as your draft. It is now refused with a message asking for the address of the site you want read.

​

Integrations

• No more surprise Google disconnects: Google Analytics and Search Console connections were sometimes disabled after a temporary token refusal even though your access in Google was untouched. A refused token is now refreshed and retried; only a genuinely revoked grant disconnects.

​

Content editor and WordPress

• One actions menu: The editor header is now two things: an actions menu (download, publish to WordPress, push to GitHub, and so on) and Save. It no longer wraps onto a second line on narrow screens.
• Editor images land in your media library: Images uploaded in the content editor are copied into your WordPress media library on publish, so posts no longer carry images that could not load.
• Media library paths accepted: A site-relative path such as /wp-content/uploads/… is accepted where an image URL is expected, and URL errors now say which field is wrong.

​

Content editor

• Version history: Every content draft keeps a history. Open History to see each version with its author (you or an agent), compare any two, and restore an earlier one. A checkpoint is kept after every 30 seconds of editing, up to 200 versions per draft, and a restore is itself a new version so it can be undone.
• Agents edit passages instead of retyping: An agent asked to change one paragraph now edits just that passage. The rest of your draft is left exactly as written.

​

Schedules

• Run now on every schedule: Run now, with its credit estimate and confirmation, is available on every schedule card, not only AI Visibility. History lists every schedule, including the gap detectors.
• New keywords are ranked promptly: Keywords added since the last check are ranked on the next sweep (within about fifteen minutes) instead of waiting until the whole project is due again. Only the new keywords are checked, so adding one keyword costs one check.

​

Agents and integrations

• Outreach agents record prospects and send email: Outreach agents can add prospects to your link-building activity sheet and send outreach from a connected mailbox, instead of handing the task back to you as blocked.
• Publish to GitHub and WordPress from a task: Publishing agents now have working GitHub and WordPress publishing tools whenever the project has that integration connected.
• Connecting GitHub is one flow: Choose the account, then the repository, in one two-step flow with a real Back button. The integration card names the connected account and lets you disconnect it.
• False claims are corrected in the thread: If an agent says it published, emailed or updated something it did not actually do, the correction appears right where you read the reply.

​

Tasks

• Your view survives a refresh: Filter, sort and group on the Tasks page are remembered per company across reloads.

​

Agents

• Project memory: Agents now keep notes per project (the site’s voice, preferences, what they are midway through), shared with every agent working that project and kept separate from other projects.
• See the steps behind the status line: The live “working” indicator on a task or chat gains a chevron that opens the steps the agent has taken so far.

​

Custom dashboards

• Every data widget has options: Use the gear on a widget to choose which ranking bands to show, which traffic figures to include, or a narrower window for a trend chart. Placeholder data in the builder now clearly reads as placeholder.

​

Projects

• A project’s domain locks once data is collected: Changing the domain after rankings, traffic or AI-visibility data exist would plot two different websites on one trend line, so it is now refused. For a rebrand, start a new project; the old one keeps its history.

​

Opportunities

• Competitor content gaps join New content: Topics your competitors rank for and you don’t now appear in the Content tab’s New content bucket, ranked by score alongside the Search Console suggestions, instead of in a separate section.
• Directories grouped separately in competitor suggestions: Marketplaces and directories (the kind of site AI assistants cite often) are listed below real competitor candidates with a note on what adding one costs, so you don’t accidentally exclude your best AI-citation finding.

​

Notifications

• No replayed toasts: Old notifications no longer pop up again on every page load.

​

Opportunities

• Three automatic gap detectors: Backlinks, Content and AI Citations are now fed by scheduled detectors that need no agent: sites that link to your competitors but not you, topics competitors rank for that you don’t, and sources that AI assistants cite for your rivals but not for you. Technical and On-page still come from audit agents.
• AI Citations replaces GEO: The tab is renamed and now groups findings by source (“this site cites rivals but not you”) rather than by competitor, so each finding is a concrete outreach target.
• Scored findings, won badges, one-click tasks: Every finding carries a 0-100 score, low-scoring ones are collapsed by default, a Won badge marks a domain that now links to you, and any finding converts to a prefilled task. When a tab can’t run yet, it names the exact missing prerequisite and links to where to fix it.

​

Project overview

• Site authority against competitors: A Site authority card shows your Domain Rating (by Ahrefs) on a 0-100 ring next to each competitor’s rating. An unrated domain shows as Unrated, never 0.
• Competitor suggestions: While your competitor list is empty, the overview suggests domains that keep appearing in your AI-visibility answers; project settings offers the same suggestions as one-tap pills.

​

Billing and team

• Top-ups show what you get: The top-up dialog converts your amount to credits live (1 credit = $0.01).
• Profile photos: People are shown with their profile picture instead of initials wherever members are listed or picked.

​

WordPress publishing

• Featured image: The featured image chosen in the editor is uploaded to your site and set on the post.
• SEO metadata reaches the site: SEO title, description, robots and the rest of the editor’s SEO panel are written to the post and mapped to whichever SEO plugin your site uses.
• Schedule a publish date: Set a future date and WordPress publishes the post itself at that time. Publishing still goes through an approval first; the date only decides when it lands once approved.

​

Agents

• Live status instead of a typing indicator: Chat and task discussions now show the agent’s current step (“Opening task TES-3”, “Updating memory”, “Reading page…”) while it works, so a long run no longer looks like a hang.
• Preset agents are editable: You can edit a preset agent’s instructions in the hire wizard or its Config tab and it stays linked to its preset. The clone-as-custom workaround is gone.
• Chat shows every agent’s history: The agent list and the conversation are one panel; expand an agent to browse its past chats while talking to another.

​

Notifications

• Toasts are back, filtered: Important events pop up as a toast with a close button, including on the task page. Everything still lands in the dock and the Inbox.
• Failures say why, and nobody’s tasks reach admins: A failed-run notification now includes the reason. Tasks that an agent created and assigned to itself, with no human owner, notify the company admins instead of no one.

​

Content editor

• No more lost work: Links get a bubble to view, edit or remove them; Insert image works; a failed save is shown and retryable instead of reading as saved; tables, checklists and underline survive editing; and the editor warns before you leave with unsaved changes.

​

Reports

• Date-range filter everywhere: Every report page and the project overview has one date control with presets and a custom from/to calendar. The selection lives in the URL, so it survives a reload and can be shared.

​

Content planner

• Named boards: A project can keep several content plans (a pillar map, a quarter’s calendar, a gap board), switched from a dropdown that also creates, renames and deletes them.
• Zoom, fit and tidy layout: Zoom with the wheel or the +/- buttons, Fit frames the whole board, and Auto arrange lays topics out as a clean tree and fits them on screen.

​

Agents

• Run timeline: A run’s page shows where its time went: prompt preparation, each model call and each tool call with durations, slowest step highlighted.
• Configuration autosaves: The agent Configuration card saves as you type, like the rest of the Config tab; the Save changes button is gone.
• One picker for who can assign an agent: Groups and people are chosen from a single searchable dropdown on the hire form and the Config tab.

​

Navigation

• Land where you left off: Opening the app returns you to the last company you were working in.

​

AI Visibility

• Run Now with a cost confirmation: Run Now on the AI Visibility schedule runs a real check, after a dialog showing the exact credit cost. Edit and Run Now are hidden until the schedule’s prerequisites are met.

​

Billing

• Balances in credits: Credit balances, history and per-run costs read as “N Credits” (1 credit = 1 cent). Subscription prices and top-up purchases, which are real card charges, still show in dollars.

​

Chat

• Full-screen diagrams and copy buttons: Diagrams in agent replies expand full-screen, every message has a copy button, and each table has its own copy and Excel buttons.

​

Reliability

• Smooth updates: After a product update the app reloads itself instead of showing a blank page.

​

Schedules

• Run history: Every scheduled check (keyword rankings, AI visibility, organic traffic, content opportunities) records each attempt. Each card shows the latest run’s status and cost, and a History sub-tab lists every past run.
• Retry after a failure: A Retry button appears on a card after a real failure.
• Estimated cost per run: Each card shows a live estimate (“~80 Credits est. / run”) before a run has happened, with the real recorded cost once one has.

​

Help

• Support button: A button in the corner of every page opens support chat and the community link.

​

Tasks

• Task list loads in pages: The list fetches fifty tasks at a time and loads more as you scroll. Filtering, search, sort and group now apply to every task, not just the ones on screen, and search reaches descriptions.
• Jump between tasks: The task title in the header is a searchable task picker, so switching tasks no longer means going back to the list.
• Warning when no agent is assigned: A task with no agent assigned says so under the assignee picker, since nothing will pick it up.
• Context meter and long descriptions: The details panel shows how full the agent’s context window is for this task, with a hint when it is getting heavy, and long descriptions get an Expand button.

​

Billing

• Credits only for those who can act: The credits card is shown to people with billing access. A workspace whose credits are all top-up reads “Reserve - no monthly plan” instead of an empty gauge.
• Filterable billing history: History splits into subscription, AI usage and SEO checks, each filter showing its own all-time total, so per-check charges are no longer buried among AI usage rows.

​

Opportunities

• Two dropdowns instead of nine pills: Status and severity filters are two dropdowns with live counts, on every Opportunities tab.

​

Agents and data

• Keyword volume and difficulty are measured: Search volume, CPC and difficulty come from DataForSEO. The AI rates only relevance to your business. Where a metric is unavailable it shows as a dash rather than a guess, and the keyword score is on the 0-10 scale it is labelled with.
• Page performance from Google: Audits measure performance, accessibility, best practices and SEO scores with Google PageSpeed Insights, and real-user Core Web Vitals (LCP, INP, CLS) from the Chrome UX Report.
• Real backlink, indexing and AI-answer data: Link agents get real backlink data and a domain’s Domain Rating from Ahrefs; indexing status comes from Google Search Console; AI Visibility reports which keywords actually trigger an AI answer.
• New keywords are checked immediately: Adding a keyword rank-checks it straight away, and only it, so the rank column fills in within moments.

​

Integrations

• A broken Google connection is disabled once, and you are told why: If Google Analytics or Search Console stops working (access removed, password changed, grant expired), the connection is disabled, the Integrations card explains the reason in plain words, and admins get an Inbox notification. It re-enables automatically if it starts working again.

​

Projects and billing

• Setup checklist: A project shows what it still needs (domain, keywords, brand names, competitors, Search Console, Google Analytics), what each unlocks, and links to where to fix it. It disappears once complete and respects features you have switched off.
• Credit balance in the sidebar: Your plan credits for the month sit above the account menu with a one-step top-up dialog, so you see the balance before a scheduled check runs it out.

​

Chat and tasks

• Export a reply as Excel, Word or Markdown: An agent reply with a table has a Download as Excel button (one worksheet per table). Right-click or long-press any agent message to copy it or save it as Word or Markdown.
• Collapsed composer: The task composer collapses to one bar with attach and send on it, and expands when you start typing.

​

Project overview

• Every report summarised: The overview carries a summary per enabled report: Tasks (completion ring, throughput), Keyword Ranking (average position, distribution), AI Visibility (score, per-assistant bars, competitors, cited domains) and Organic Traffic (trend, top queries, pages, landing pages), each with a button through to the full report.
• Organic traffic with one Google connection: The Organic Traffic report renders with only Google Analytics or only Search Console connected.

​

Keywords and AI Visibility

• See what tracking costs per month: Manage keywords is now two tabs, Add keywords and Saved, and both keywords and AI-visibility prompts show their monthly credit cost with the arithmetic spelled out. The per-run keyword cap is gone, so every tracked keyword is checked.
• Agents propose keywords: Audit agents suggest keywords for you to accept or dismiss instead of editing your tracked list. A dismissed suggestion is not proposed again.

​

Team

• Invites are emailed: Creating a team invite sends an email with the link; you can still copy the link yourself.
• Agents stay in their projects: Agents can only act on the projects they are assigned to.

​

Account

• Sign in with X: X (Twitter) joins Google and email/password as a sign-in method. Account settings show which methods are connected.

​

Projects and companies

• Logos: Upload a logo for a company and for each project; it appears on cards, in the sidebar and in the switcher.
• One-step project creation with an AI-written description: Creating a project is a single step and every feature starts enabled. A star button beside the description reads your site and writes the brief your agents receive on every task; you can edit or ignore it.

​

Team and agents

• Groups: Team gains a Groups tab for reusable sets of people. Groups carry no permissions; they make picking people faster.
• Who can assign this agent: Each agent has an allow-list of people and groups who may assign it work, on the hire form and on its Config tab. It widens who can assign beyond the reporting line.

​

Integrations

• Why a Google connection needs reconnecting: The Integrations card says why a connection needs attention (a missing consent scope, a disabled API, lost property access) and what to do about it, instead of a bare Reconnect button.

​

Design

• Neutral near-black palette: Dark mode moves to a neutral near-black palette, and the Activity page uses real tabs at full width.

​

A new platform

• Rebuilt from the ground up: Visibility 2.0 is a from-scratch rebuild at app.visibility.so. Everything below ships today; the changelog for the previous product is retired.

​

Agents

• A Head Agent for every company: Each company has one Head Agent (the SEO Manager) that designs and hires the specialist agents you need. Hire from a catalogue of presets: Keyword Research, Technical SEO Audit, On-page Audit, Content Writer, Topic Research, Competitor, Backlink Building, SEO Outreach, WordPress Publisher and more, or describe an agent and let the Head Agent propose it for approval.
• Managed AI or bring your own key: Agents run on Visibility’s managed AI, metered in credits, or on your own AI provider keys. Every run has a live transcript, and a Stop button ends a run instantly.
• Chat with any agent: Open Chat, pick an agent and talk to it directly, with images, attachments and follow-up messages while it works.

​

Tasks

• Tasks, not tickets: Create a task with a title and, optionally, a description, assignee (person or agent), priority, project, category, dates and attachments. Agents pick up assigned tasks, post updates as comments and mark them Done. Recurring tasks run on a schedule you set with day-and-time pickers in your timezone.
• Subtasks and delegation: Agents can break work into subtasks and delegate to peers, and are woken when a subtask completes.

​

Projects and reports

• One project per site: A project holds a domain, competitors, brand names, topics and location. Its sidebar carries Overview, Tasks, Opportunities, Content, Link building and Settings, plus the reports: Tasks Overview, Keyword Ranking, AI Visibility, Organic Traffic and Custom Dashboards.
• AI Visibility: Track how often your brand appears in ChatGPT, Claude, Gemini and Perplexity answers for the prompts and topics that matter, with brand presence, cited domains and competitors mentioned, tracked over time.
• Keyword Ranking: Track keywords per country and device, with volume, CPC and difficulty, average position over time and a position distribution.
• Organic Traffic: Connect Google Analytics and Google Search Console for one combined traffic report: sessions, clicks, impressions, top queries, pages, devices and landing pages.
• Opportunities: Technical, On-page, Content, Backlinks and AI Citations tabs list what to fix or pursue, each convertible into a prefilled task for the right agent.
• Custom Dashboards: Build report pages by dragging widgets onto a grid, and share them with a link.

​

Content and link building

• Content editor with WordPress publishing: Write and edit drafts in a rich editor with an SEO score panel, pull a page in from a URL, plan topics on a visual planner, and publish to a paired WordPress site through the Visibility plugin. Every publish is an approval in your Inbox.
• Link building workspace: Activities with live previews, an editable prospect sheet, tasks and a database, worked by the Backlink Building and SEO Outreach agents with connected mailboxes (Gmail or SMTP) for outreach and follow-ups.

​

Team, Inbox and billing

• Inbox and approvals: The Inbox is a notification centre with two tabs, Inbox and Approvals. Exactly two things need approval: hiring an agent and publishing to WordPress.
• Team roles and permissions: Invite teammates as admins or members, or define custom roles with a permission matrix. Agents report to people in an org chart that decides who can assign them work.
• Plans, credits and secure checkout: Companies start on a Starter trial and choose Starter, Growth or Enterprise. Paid SEO checks and managed AI draw from a monthly credit allowance; buy more credits or change plan any time through secure checkout, with a full billing history.
• Sign in and search: Sign in with Google or email and password. Ctrl+K opens universal search across tasks, agents, projects and pages, and a public status page shows platform health.