Lokalog

🌟 What's new?

You can now create Machine Translation tasks directly from Lokalise's task-creation form instead of relying on Workflows or Orders. Choose the provider and languages while keeping task creation in one place.

• Supported providers: Google and DeepL.
• The source language is automatically preselected based on your project's source language and cannot be changed.
• Target languages are limited to supported Machine Translation language pairs.
• Machine Translation tasks use Processed Words as part of the Standard AI offering.

🎯 Why it matters?

You can start Machine Translation tasks faster directly in Lokalise and choose between Google and DeepL when the required language pair is supported. This simplifies task creation and keeps the process in one place.

🌟 What's new?

AI scoring no longer runs automatically when a workflow triggers a Pro AI translation task. It now only runs when the workflow explicitly includes it.

Workflows that include AI scoring (scoring will run as before):

• Pro AI + Review task with AI scoring
• TM + Pro AI + Review task with AI scoring

Workflows without AI scoring (scoring will not run):

• Pro/Standard AI
• Pro/Standard AI + Review task
• TM + Pro/Standard AI
• TM + Pro/Standard AI + Review task

If you rely on AI scoring in your workflows, make sure you're using one of the templates above. No other action needed.

🎯 Why it matters?

• AI scoring now runs intentionally, not silently in the background on every task.
• You only get scoring when you need it — giving your team cleaner control over how AI is used in your localization process.
• No pricing change.

🌟 What's new?

• Google's Gemini 3.5 Flash joins GPT and Claude as a selectable AI model in Lokalise.
• You can now choose Gemini 3.5 Flash directly from your AI Profile settings — no extra setup required.
• All three models are available at no additional cost.

🎯 Why it matters?

• You now have three leading AI models to choose from, giving you more flexibility to match the right engine to your content and language pairs.
• Whether you're already happy with GPT or Claude, or want to explore what Gemini 3.5 Flash brings to your workflows, the choice is yours — all in one place.
• Another step in Lokalise's commitment to giving you access to best-in-class AI, regardless of which model leads for your use case.

🌟 What's new?

• Bulk push for nested references — you can now push content with deeply nested references in a single operation, with control over whether referenced entries are included. No more pushing entries one by one.
• Scheduled imports now include references — automated imports now behave consistently with manual ones, so nothing gets left behind when running on a schedule.
• Faster and more reliable imports — import times are noticeably shorter, and jobs no longer fail silently. Errors now surface clearly so you know exactly what to fix.

🎯 Why it matters?

If you manage complex Contentful content models, you'll feel the difference right away — less manual work, fewer surprises, and imports that complete faster and more predictably.

🌟 What's new?

You can now access Lokalise audit log events programmatically via a new public REST endpoint — GET /v1/audit-logs — for the first time.

• Audit events returned in OCSF 1.3.0 format, natively supported by IBM QRadar, Splunk, Microsoft Sentinel, and Datadog.
• Cursor-based pagination for reliable, resumable polling.
• Filter by date range, event type, project, or user.
• Part of the new api.lokalise.com/v1/ path — all future public endpoints will live here. Existing /api2 integrations are completely unaffected.

Requires team admin access and a plan that includes Audit Logs.

🎯 Why it matters?

• Until now, audit logs were only viewable in the Lokalise UI — no way to export or pipe them into external security tooling.
• OCSF 1.3.0 compatibility means no custom parsing or data mapping — plug straight into your SIEM.
• Unlocks automated compliance monitoring, real-time security alerting, and audit trail retention in your own systems.

🌟 What's new?

Lokalise now sends two security notifications to your original email address whenever an email change is requested:

• On request — you're notified the moment a change is initiated.
• On confirmation — a follow-up email confirms once the change is complete.

Both include clear instructions to contact Lokalise support if the change wasn't you.

🎯 Why it matters?

• Your account email is a critical access credential — you're now always in the loop if someone tries to change it.
• For teams managing multiple users and sensitive localization data, this is a meaningful step toward tighter account security and auditability.

🌟 What's new?

Two new configuration options are now available in Workflows:

• Post TM status for "Apply TM" steps — when setting up an Apply TM step, you can now choose what status gets assigned to TM-matched translations automatically: mark them as Reviewed, Unverified, or keep the current status.
• Exclude reviewed translations from review tasks — when configuring a Review task step (including AI Scoring review tasks), you can now exclude translations already marked as Reviewed from the task scope. Reviewers only see segments that actually need attention.

🎯 Why it matters?

• TM matches can now flow through your workflow fully hands-off: automatically filled, flagged as reviewed, and skipped in downstream review steps — so your team's time goes where it counts.
• Previously, TM-matched translations still appeared in review tasks, creating unnecessary work and reducing the cost-saving value of TM. That friction is now gone.
• This makes Workflows a more powerful tool for cutting review costs and building leaner localization processes.

🌟 What’s new?

Lokalise Glossary Guard, the tool that automatically checks and fixes your glossary CSV files before you upload them to Lokalise — is now available as a browser-based web app, in pre-release.

• Simply visit lokalise-glossary-guard.web.app, upload your glossary CSV file, and download the fixed version — no terminal or installation required.
• The tool automatically detects and resolves common glossary issues such as wrong separators, encoding problems, duplicate entries, and formatting inconsistencies.
• This is a pre-release version, so expect ongoing improvements based on early feedback.

🎯 Why it matters to you?

• Getting your glossary files upload-ready used to mean either fixing issues by hand or running a command-line tool — neither ideal if you're not a developer.
• This web app makes Glossary Guard accessible to everyone on your team, no technical setup needed.
• You can now validate and clean glossary files in seconds, reducing upload errors and keeping your localization workflows running smoothly.

🌟 What’s new?

Translation Memory just got smarter about what counts as a validated translation:

• Previously, when AI or machine-translated content went through a human review task, only translations that were manually edited by the reviewer were saved to TM.

• Now, any AI/MT translation that a reviewer marks as Reviewed (even without edits) is automatically saved to TM under the reviewer's name.

• This applies regardless of how the reviewed status is set: clicking the Reviewed icon, using a keyboard shortcut, bulk actions, or saving within the review task.

🎯 Why it matters to you?

If you run AI or machine translation through a human review step, you've already been doing the hard work of validating those translations, now Lokalise captures that work:

• Grow your TM faster, without requiring reviewers to make edits just to trigger a save.

• Reuse more validated content in future projects, improving TM leverage and reducing repeated review of already-approved translations.

• Build a higher-quality TM over time, since every entry is traceable to a human reviewer who signed off on it.

🌟 What’s new?

Analytics has a new section: Translation Quality. It gives you a clear picture of how much post-editing your reviewers are doing and where quality issues originate across your localization workflows. The new reports include:

• Post-edit rate & average edit distance, understand how heavily AI or machine translations are being corrected
• Review effort distribution, see how effort is spread across languages, workflows, and contributors
• Workflow & language quality patterns, identify which setups produce the most reliable translations
• Translator and reviewer trends, track individual and team-level quality signals over time

One important detail: cases where reviewers had to fully retranslate or create new translations from scratch are tracked separately. They appear in the broader review-effort visualizations but are excluded from post-edit rate and edit distance calculations, so your quality metrics stay accurate and aren't skewed by retranslation events.

This release also brings a refreshed Analytics experience overall: improved navigation, OTA reporting now available directly in Analytics and updated copy throughout.

🎯 Why it matters to you?

Until now, Analytics gave you strong visibility into volume and throughput, but not into whether translations were actually good. This release fills that gap:

• Know if your translations are ready, spot patterns that signal quality problems before they reach production
• Understand real reviewer workload, see how much post-editing work reviewers are actually doing, not just how many tasks they completed
• Find the workflows that scale, identify which language/workflow combinations are efficient and which need attention
• Trace quality issues to their source, pinpoint whether problems stem from a specific language, contributor, or workflow step