Computer-Assisted Translation Tools: The Working Translator's Field Guide
Computer-assisted translation tools are segment-based editors that pair translation memory, term bases, and inline QA so working translators produce faster and
Working translators and production leads live inside the segment editor. That editor is the computer-assisted translation (CAT) tool — the surface where source files become auditable, certifiable bilingual output.
Key points
- A CAT tool is the segment-level production surface where source, translation-memory match, term-base hit, and inline QA warning co-exist in one view — alt-tabbing to a glossary is the cost signal of a bad setup.
- Translation memory leverage uses a 75% fuzzy-match floor, with bands at 75–84%, 85–94%, 95–99%, 100% exact, and 101% in-context, each discounting the $0.10–$0.25 per-source-word base rate.
- Professional translators cluster around 7 CAT tools in 2025: SDL Trados Studio, memoQ, Phrase, Smartcat, Smartling, OmegaT, and Wordfast, with MateCat and Lokalise filling adjacent lanes.
- A CAT tool owns segment, TM, term base, QA, and bilingual export; a TMS owns routing and analytics above it; an MT or LLM engine generates first drafts beneath it.
- Certified lanes for USCIS, courts, and registries priced at $0.10–$0.25 per source word or $25–$150 per standard page of about 250 words remain desktop-workbench territory because of offline file control and chain-of-custody requirements.
What is a computer-assisted translation tool?
A computer-assisted translation tool is a segment-based editor that splits a source document into translation units, pairs each unit with translation-memory matches and term-base hits, and runs inline quality checks so a human translator produces a bilingual file faster and more consistently than typing in a word processor. The tool is the one screen where source, TM match, term-base hit, and inline QA warning co-exist. Alt-tabbing to a glossary is the cost signal of a bad CAT setup.
Buyers and translators often confuse the CAT tool with three adjacent things. The disambiguation table below resolves the confusion before the rest of the page builds on it.
| Term | What it is | When you need it |
|---|---|---|
| CAT tool | Segment editor with TM, term base, and inline QA | Every project a human translator touches |
| Translation memory (TM) | Database of past approved source + target segment pairs | Any project with repeated or similar content |
| Translation management system (TMS) | Project-layer platform for routing, connectors, vendors, analytics | Agency programs with multiple files, locales, or vendors |
| Machine translation (MT) engine | First-draft generator — NMT or LLM — populating target cells | Volume content with human post-edit downstream |
For a buyer-side selection view that complements this field guide, see What a CAT Tool Does and How to Choose One.
How is a CAT tool different from machine translation?
A CAT tool is the human translator's editing surface; a machine-translation engine is a first-draft generator that may pre-populate the target cell inside that surface. The translator owns every segment, approves TM matches, and signs off on terminology before export. The engine proposes; the translator and the CAT tool dispose.
How is a CAT tool different from a translation management system (TMS)?
A CAT tool owns the segment; a TMS owns the project. The CAT editor is where TM leverage, term-base enforcement, and QA warnings surface in one view. The TMS routes jobs, syncs repositories, tracks vendors, and reports analytics around that editor. Agencies evaluating the layer above the CAT seat should read How to Choose a Translation Management System for Your Agency.
What are the core features every CAT tool provides?
Every production-grade CAT tool provides 6 core features: segment-based editing of source and target, a translation memory with fuzzy and exact match leverage, a term base with real-time enforcement, inline QA checks on tags and numbers, bilingual file formats such as XLIFF for interchange, and project packaging for multi-translator handoff.
- Segment-based editing — two-column grid, source locked, target typed inline.
- Translation memory leverage — fuzzy, exact, and in-context matches pre-filled against a 75% floor.
- Term base enforcement — real-time flagging as the translator types.
- Inline QA — tag mismatches, number drift, forbidden terms, punctuation drift, segment-length anomalies.
- Bilingual file formats — XLIFF 1.2/2.0, SDLXLIFF, MQXLIFF, TXLF, plus TMX and TBX for exchange.
- Project packaging — bundling source, TM, TB, and settings for multi-translator handoff.
flowchart TD
A[Source file: DOCX, IDML, PDF, XLIFF, JSON] --> B[Segmenter splits into translation units]
B --> C[CAT editor two-column grid]
C --> D{TM query per segment}
D -- 101 percent --> E[In-context match pre-fills target]
D -- 100 percent --> F[Exact match pre-fills target]
D -- 75 to 99 percent --> G[Fuzzy match pre-fills, translator edits]
D -- below 75 percent --> H[No match, translator writes target]
C --> I[Term base enforcement flags deviations]
C --> J[Inline QA on tags, numbers, punctuation]
E --> K[Bilingual export: XLIFF, SDLXLIFF, MQXLIFF]
F --> K
G --> K
H --> K
J --> K
K --> L[Target file rebuild with original formatting]
How does segment-based editing work?
The CAT tool splits the source file at sentence or paragraph boundaries into translation units, locks source formatting into inline tags, and presents a two-column grid where the translator types the target next to each source segment without touching the surrounding layout. The segmenter preserves tables, bullets, and inline styles as tags, so a bold run or an InDesign style sheet survives round-trip without manual re-formatting. Software that handles this cleanly across InDesign and Word inputs is covered Software That Handles Document Translation Without Losing Formatting.
How does translation memory leverage work inside a CAT tool?
Translation memory stores every approved source-and-target segment pair from past projects; the CAT tool queries that database on every new segment and pre-fills the target cell when a match crosses a client-agreed threshold. The typical floor is 75% for fuzzy matches, 100% for exact matches, and 101% for in-context matches where the preceding and following segments also match. See How CAT Tools Use Translation Memory to Speed Up Your Work for the mechanics in production, and Choosing Translation Memory Software: What to Compare Before Buying for the buying criteria.
What do the TM match bands mean in production?
Match bands map directly to per-word economics. 101% in-context matches go through at a nominal fee, 100% exact matches at 10–30% of the full rate, 75–99% fuzzy at 50–70%, and no-match segments at the full $0.10–$0.25 per source word.
| Match band | What it means | Share of full rate | Effective per-word cost (base $0.10–$0.25) |
|---|---|---|---|
| 101% in-context | Segment and both neighbors match | 0–10% | $0.00–$0.025 |
| 100% exact | Segment matches, context differs | 10–30% | $0.01–$0.075 |
| 95–99% fuzzy | One token or tag differs | 50–60% | $0.05–$0.15 |
| 85–94% fuzzy | Minor rewrite | 60–70% | $0.06–$0.175 |
| 75–84% fuzzy | Partial reuse, significant edit | 70% | $0.07–$0.175 |
| No match (<75%) | Full translation | 100% | $0.10–$0.25 |
How does term base enforcement work?
A term base stores approved renderings of product names, feature names, and regulated terminology; the CAT tool flags deviations in real time as the translator types. A mis-rendered drug name or trademarked term never reaches the LQA gate. In regulated verticals that cost of a missed term is why desktop workbenches dominate — see Medical Translation Software: What Professionals Actually Use.
What do inline QA checks catch before export?
Inline QA checks catch 5 defect classes before the file leaves the editor:
- Missing or reordered inline tags that would break formatting on rebuild.
- Number and date mismatches between source and target segments.
- Forbidden-term usage against the term base.
- Double spaces, trailing punctuation, and quotation-mark drift.
- Segment-length anomalies that signal untranslated or over-translated text.
Which bilingual file formats does a CAT tool read and write?
A production-grade CAT tool reads and writes XLIFF 1.2 and 2.0 as the open interchange standard, plus vendor-specific bilingual formats such as SDLXLIFF (Trados), MQXLIFF (memoQ), and TXLF (Wordfast). TMX handles translation-memory exchange and TBX handles term-base exchange.
- XLIFF 1.2 / 2.0 — OASIS open standard for bilingual segments in-flight.
- SDLXLIFF — Trados's XLIFF dialect; the desktop market default.
- MQXLIFF — memoQ's XLIFF dialect for server-based sharing.
- TXLF — Wordfast's bilingual format.
- TMX — translation-memory exchange across tools.
- TBX — term-base exchange across tools.
For web content shipped as JSON or XLIFF strings, see Software Built for Translating Web Content Without Losing Markup.
Which computer-assisted translation tools do professional translators actually use?
Professional translators cluster around 7 CAT tools in 2025: SDL Trados Studio as the desktop market leader, memoQ as the agency workhorse, Phrase as the cloud-first contender, Smartcat and Smartling as browser-based platforms, OmegaT as the open-source option, and MateCat and Wordfast as lower-cost alternatives still common in freelance work.
| Tool | Deployment | Bilingual format | Best fit | Pricing band (US, 2025) |
|---|---|---|---|---|
| SDL Trados Studio | Desktop | SDLXLIFF | Freelance + agency, certified lanes | $400–$900 perpetual |
| memoQ | Desktop + server | MQXLIFF | Agency, multi-translator packages | $620–$900 perpetual; server licenced |
| Phrase | Cloud | XLIFF | Software strings, continuous content | $27–$80 per user per month |
| Smartcat | Cloud | XLIFF | Agency marketplaces, freelance | Free seat; paid tiers |
| Smartling | Cloud | XLIFF | Web localization, continuous content | Enterprise contract |
| OmegaT | Desktop | XLIFF, TMX | Cost-sensitive freelance | Free, open source |
| MateCat | Cloud | XLIFF | Freelance, light projects | Free |
| Wordfast | Desktop + web | TXLF | Freelance, Word-centric lanes | $460 perpetual (Pro) |
| Lokalise | Cloud | XLIFF, JSON | Software strings | $120+ per month, team pricing |
A broader buying view for agencies is Business Translation Software Worth Paying For. Specialist and regional tools sit alongside this list — see Kiwi CAT Tool: What It Does and Who It Suits.
What does SDL Trados Studio do and who uses it?
SDL Trados Studio is the desktop CAT market leader used by individual freelancers and large agencies for high-stakes documents where offline file control matters. It owns the SDLXLIFF bilingual format and ships with the deepest TM, term base, and QA integrations in the category. Full feature and cost detail sits Trados as a CAT Tool: Features, Costs, and Honest Trade-offs.
What does memoQ do and who uses it?
memoQ is the agency-side workhorse, strongest on multi-translator project packages, server-based TM sharing, and the MQXLIFF bilingual format. The paid translator pro licence adds agency-grade features over the free edition — see MemoQ Translator Pro: What the Upgrade Adds Over the Free Version. The browser variant is covered What MemoQWeb Offers Translators Working in a Browser, and the full comparison What memoQ Does and How It Compares to Other CAT Tools.
What does Phrase do and who uses it?
Phrase is a cloud-first CAT platform used heavily on continuous software-string work, with repository connectors, in-context editing, and the Phrase AI pre-translate layer that scores engine output per segment before a human post-edits. Day-to-day features sit Phrase CAT Tool: Features Translators Use Every Day. If Phrase is not the right fit, review Phrase Alternatives Worth Considering for Your CAT Workflow.
What does OmegaT do and who uses it?
OmegaT is the open-source CAT tool used by cost-sensitive freelancers and translators who want local file control without a licence fee. It handles XLIFF and TMX cleanly but ships without the server-side TM sharing and QA depth of commercial platforms. The strengths and the gaps are itemised OmegaT for Professional Translators: What It Does and What It Lacks.
What do Smartcat, Smartling, Lokalise, and MateCat offer?
These four are browser-based CAT platforms that fold the TMS and the segment editor into one interface. They fit continuous content and software strings better than certified-document lanes, where offline file control is a chain-of-custody requirement.
- What Smartcat Offers as a CAT Tool and Where It Falls Short — marketplace model, free translator seat.
- What Smartling Offers and How Its CAT Tool Fits a Translation Workflow — enterprise web localization.
- Is Lokalise a Viable CAT Tool for Localization Translators? — software-string focus.
- MateCat Reviewed: Features, TM Handling, and Fit for Freelancers — free tier that queries MyMemory Translation Memory: How It Works and When to Use It.
How does a CAT tool fit into the full translation production stack?
A CAT tool sits in the middle of a 3-layer stack. The TMS above routes jobs, syncs repositories, and holds analytics. The MT or LLM engine below generates first drafts per segment. The CAT tool is the segment-level production surface where TM matches, term-base hits, and inline QA warnings converge in one view the translator never alt-tabs away from.
For how agencies run the whole pipeline, see Localization Management: How Translators and Agencies Run Multi-Locale Programs.
What does the CAT editor own vs. the TMS vs. the MT engine?
The CAT editor owns segment, TM, term base, inline QA, and bilingual export. The TMS owns routing, connectors, vendor assignment, analytics, and project lifecycle. The MT engine owns first-draft generation per segment and feeds its output into the CAT target cell for human post-editing.
| Function | CAT editor | TMS | MT / LLM engine |
|---|---|---|---|
| Segment creation and editing | Owns | — | — |
| Translation memory query | Owns | Stores shared TM | — |
| Term base enforcement | Owns | Stores shared TB | — |
| Inline QA | Owns | Aggregates reports | — |
| Bilingual export | Owns | Packages for delivery | — |
| Job routing + vendor assignment | — | Owns | — |
| Repository connectors | — | Owns | — |
| First-draft target generation | Consumes | — | Owns |
| Confidence scoring per segment | Displays | Reports | Owns |
Workflow context across the layers is How Translation Management Software Fits Into Your Workflow.
How do CAT tools interoperate across tool boundaries?
CAT tools interoperate through 3 open formats: XLIFF for bilingual segments in-flight, TMX for translation-memory exchange across tools, and TBX for term-base exchange. A project started in memoQ finishes in Trados when XLIFF round-trips cleanly without segment loss or tag damage.
The interoperability map in practice:
- XLIFF 1.2 — accepted by Trados, memoQ, Phrase, Smartcat, OmegaT, MateCat, Wordfast.
- XLIFF 2.0 — accepted by Trados, memoQ, Phrase, OmegaT; cloud platforms lag.
- SDLXLIFF — native in Trados; imported by memoQ with occasional tag loss.
- MQXLIFF — native in memoQ; imported by Trados with manageable fidelity.
- TMX 1.4 — universal round-trip across every production-grade CAT tool.
TBX — accepted broadly but with vendor-specific extensions that need manual mapping.
Cloud-TMS to desktop-CAT round-trips are typical on Crowdin-run projects — see Crowdin Translation: How Translators and Agencies Run Projects Inside Crowdin.
Where does AI pre-translate fit inside the CAT tool?
AI pre-translate populates the target cell with an LLM or NMT draft before the translator touches the segment. The CAT tool scores confidence per segment, routes low-confidence segments to human post-edit, and holds the output for LQA sampling at $40–$90 per 1,000 words against an MQM-style scorecard aligned to ISO 17100. The full first-draft-plus-human-certification shape is AI Localization: How Translators and Agencies Run the First-Draft + Human-Certification Workflow, and the Phrase-specific implementation What Phrase AI Does and Whether It Fits Your Workflow.
How do you choose a CAT tool for your own translation work?
Choose a CAT tool against 6 criteria: file-format breadth across your real client inputs, translation-memory and term-base interoperability through XLIFF and TMX, inline QA depth, pricing model, fit for your highest-value lane, and vendor lock-in risk measured by how cleanly projects export.
- File-format breadth across your real client inputs.
- TM and term-base interoperability through XLIFF and TMX.
- Inline QA depth — tag, number, terminology, punctuation, length.
- Pricing model — perpetual, subscription, or open source.
- Fit for your highest-value lane — certified, continuous software, or marketing.
- Vendor lock-in risk — tested by round-tripping a 20-segment file.
The companion buyer checklist is What a CAT Tool Set Includes and How to Pick One.
Which file formats do real translation clients send?
Match the CAT tool to the file formats your clients actually send. InDesign IDML and PDF for marketing agencies, DOCX and PPTX for corporate clients, XLIFF and JSON for software clients, and scanned PDFs with OCR for USCIS and legal filings.
Does the CAT tool round-trip XLIFF and TMX cleanly?
Test XLIFF and TMX round-trips before committing. Import a bilingual XLIFF from one tool, translate 20 segments, export, and re-import in the original tool. Segment loss, tag damage, or TM mismatches on round-trip signal vendor lock-in you will pay for later.
What does a CAT tool cost per translator per year?
CAT tool pricing in the US falls into 3 bands in 2025: perpetual desktop licences at $400–$900 one-time plus annual upgrades (Trados, memoQ), subscription cloud seats at $25–$80 per user per month (Phrase, Smartcat, Lokalise), and free open-source options (OmegaT, MateCat) with zero licence cost but no vendor support.
| Pricing model | Representative tools | Year-one cost per seat | Trade-off |
|---|---|---|---|
| Perpetual desktop | Trados, memoQ, Wordfast | $400–$900 + optional support | Capital outlay; full offline control |
| Cloud subscription | Phrase, Smartcat, Lokalise, Smartling | $300–$960 per user per year | Lower entry; vendor dependency |
| Open source | OmegaT, MateCat | $0 | No vendor support; thinner QA |
Does a CAT tool fit a certified-translation lane?
Certified-translation lanes require offline file control, named-translator attestation inside the bilingual export, and a documented chain of custody that cloud multi-tenant platforms struggle to prove. Desktop CAT workbenches remain the default for certified filings for USCIS, courts, and registries priced at $0.10–$0.25 per source word or $25–$150 per standard page of about 250 words. The cloud-vs-desktop trade-off for working translators is weighed Cloud-Based CAT Tools Compared for Working Translators.
How do CAT tools change the economics of a translation project?
CAT tools change project economics on 3 axes. Translation-memory leverage discounts repeated and fuzzy-matched segments against the base $0.10–$0.25 per source word. Review at $0.04–$0.08 per word compounds that discount across the file. LQA sampling at $40–$90 per 1,000 words becomes the only quality gate that keeps the discount defensible.
How does the 75% fuzzy-match threshold set the discount grid?
The 75% fuzzy threshold is the floor below which TM matches are not leveraged. Above it, bands at 75–84%, 85–94%, 95–99%, 100% exact, and 101% in-context each carry a progressively steeper discount against the full per-word rate.
How does a mid-size program scale with CAT leverage?
A mid-size multi-locale program covering 5 languages and 50,000 words costs $25,000–$75,000 depending on TM leverage, turnaround speed, and certification requirements. The spread between the floor and the ceiling is almost entirely a function of how deeply the CAT tool's TM reuses prior output.
| Scenario | TM leverage profile | Per-locale cost (10,000 words) | Program total (5 locales) |
|---|---|---|---|
| Cold start, no TM | 100% no-match | $10,000–$15,000 | $50,000–$75,000 |
| Second release | 30% leveraged, 70% no-match | $7,000–$10,000 | $35,000–$50,000 |
| Mature TM, continuous content | 60% leveraged, 40% no-match | $5,000–$7,500 | $25,000–$37,500 |
What are the limits and failure modes of CAT tools?
CAT tools fail in 4 predictable modes.
- Tag damage on round-trip between desktop and cloud editors, especially when SDLXLIFF or MQXLIFF cross tool boundaries without lossless mapping.
- TM pollution when unapproved segments enter the memory and resurface as high-confidence matches on later projects.
- Term-base drift when multiple projects share one base without governance, and conflicting approved renderings accumulate.
- Over-reliance on fuzzy matches that read fluent in context but mistranslate the segment the client actually paid for.
What does a CAT tool look like at the segment level?
A specimen of the editing grid as the translator sees it — the surface where every other feature on this page converges in one view.
Frequently asked questions
What computer-assisted translation tools do professional translators actually use?
Professional translators most use SDL Trados Studio, memoQ, Déjà Vu, Wordfast, and Smartcat as their primary CAT tools. These platforms share core features — translation memories (TMs), termbases, and segment-locking — but differ in price, collaboration features, and file-format support. Trados dominates agency workflows and is often required by large LSPs, while memoQ is favored for its project-management depth. Cloud-based tools like Smartcat and Phrase (formerly Memsource) are gaining ground for real-time team collaboration.
How is AI translation different from computer-assisted translation (CAT)?
AI translation (machine translation, or MT) generates a full translated text automatically, while computer-assisted translation (CAT) is a human-driven environment that presents segments for a translator to write or edit with tool-assisted support. CAT tools store approved translations in a translation memory and enforce terminology via a termbase, keeping the human in control of every decision. MT can be plugged into a CAT tool as a suggestion engine, but the two are architecturally distinct: one automates output, the other augments the human translator's productivity.
What is the main purpose of computer-assisted translation?
The main purpose of a CAT tool is to increase translator consistency and speed by recycling previously approved translations through a translation memory. When a new source segment matches or closely resembles a stored segment, the tool proposes the earlier translation as a fuzzy or exact match, reducing rework. For certified translation work, this is especially valuable: standard legal boilerplate, recurring government phrases, and document headers often repeat across files, and a well-maintained TM can cut active translation time on repeat content by 30–70%.
What are examples of CAT tools?
Well-known CAT tools include SDL Trados Studio, memoQ, Wordfast Pro, Déjà Vu X, Smartcat, Phrase (Memsource), Lokalise, and OmegaT (open-source). Each segments source documents, stores approved translations in a TM, and manages project-specific termbases. Some, like Trados and memoQ, are desktop-first with cloud add-ons; others, like Smartcat and Lokalise, are fully cloud-native. OmegaT is free and popular among freelancers working on Linux or with tight budgets.
What are CAT tools for translators?
CAT tools are specialized software environments that break a source document into segments and present each one for a translator to translate, with real-time suggestions from a translation memory and termbase. They are not translation engines — the human writes every target-language segment, but the tool surfaces relevant prior work and flags terminology deviations. For certified translation producers, CAT tools also preserve document formatting and create auditable records of translator decisions, which supports quality-assurance workflows and revision tracking.
What are the disadvantages of machine translation?
Machine translation frequently produces fluent-sounding but factually or legally incorrect output, which is its most serious disadvantage for certified translation work. MT systems can mistranslate names, dates, document numbers, and legal terms in ways that are hard to spot on a surface read. They lack contextual judgment about register, jurisdiction-specific phrasing, and cultural conventions. Post-editing MT to certification standard often takes nearly as long as translating from scratch when the source document is complex or poorly OCR'd, erasing much of the speed benefit.
What is one major risk of machine translation?
The biggest single risk of machine translation is confident mistranslation — the system produces a grammatically smooth sentence that carries the wrong meaning, and the error goes undetected. In certified translation, this can invalidate a legal document, cause an immigration application to be rejected, or create liability for the certifying translator or agency. Unlike an obvious garbled output that prompts review, a fluent-but-wrong sentence can pass a quick scan and reach the end client uncorrected.
Which is better, Trados or memoQ?
Neither is universally better — the right choice depends on your workflow. Trados Studio has the largest market share among agencies and is the de facto standard requested by large LSPs, making it the safer choice if agency compatibility is a priority. memoQ offers a more intuitive project-management interface and is often preferred by in-house teams or agencies managing complex multilingual projects. Pricing is comparable for professional licenses (roughly $695–$900/year for a single translator seat), and both support XLIFF, bilingual Word, and most major formats used in certified translation.
Can I use Smartcat for free?
Yes, Smartcat offers a free tier that includes unlimited translation memory, basic CAT editor access, and a marketplace for finding work. The free plan covers most core CAT features a solo freelance translator needs, making it a practical entry point. Paid plans (starting around $50–$120/month depending on team size) unlock advanced automation, MT engine integrations, API access, and priority support. For certified translation agencies managing multiple translators, the paid tiers add workflow controls that the free plan lacks.
Which is the best cat tool for translation?
There is no single best CAT tool — the optimal choice depends on your volume, collaboration needs, and client requirements. SDL Trados Studio is the industry-standard choice for agency compatibility and file-format breadth. memoQ is widely preferred for project management and team use. Smartcat suits budget-conscious freelancers and small agencies with its free tier. For certified translation production at scale, the best tool is the one that integrates cleanly with your QA process, supports your file types, and matches the TM format your clients expect.
Related
- Content Localization: How Translators and Agencies Produce It
- Continuous Localization: How Translators and Agencies Run the Delivery Model
- Crowdin Translation: How Translators and Agencies Run Projects Inside Crowdin
- Define Localization: What the Term Means in Professional Translation
- Define Transcreation: What It Means Inside a Translator's Workflow
- E-commerce Localization: What Translators and Agencies Actually Deliver