Keep every word your business publishes clean.
One API and a family of tools that catch profanity, hate speech and obfuscated slurs in text, web forms, files, databases and Microsoft Office documents, then clean them if you ask.
Great service, but the delivery driver was a complete a$$hole about the parking and I'm p!ssed off.
Great service, but the delivery driver was a complete ******* about the parking and I'm ****** off.
Wherever people write, something eventually needs catching
Twelve situations where profanity, hate speech or disguised abuse causes real damage, and how METIS HALO stops it before it reaches a page, a record or a client. Open any item to read the reasoning and the recommended set-up.
Churches, mosques, temples and charities invite the public to write: prayer requests, testimonies, condolence books, comments on sermons. One abusive or blasphemous post on a public page damages trust with the whole congregation, and volunteers cannot moderate around the clock.
Validate every form field before it is published, so offensive text never reaches the page, and scan the newsletter or bulletin in Word before it goes out. Existing comment tables can be checked in one pass.
Agents type thousands of free-text notes a day under pressure. Frustration, jokes or a customer's own language end up in CRM records that are later exported, shared with clients or disclosed under a subject access request. That is a reputational and compliance risk.
Scan notes at the moment they are saved through the API, run a nightly database scan of the notes and survey tables, and check exported spreadsheets with the Excel add-in before they leave the building.
Reviews sell products, but unmoderated reviews carry insults, competitor slurs and obfuscated swearing that filters based on simple word lists miss. Holding every review for manual approval slows the site down and costs staff time.
Call the API when a review is submitted: clean text is published straight away, flagged text is held or cleaned automatically. Leet-speak, spaced-out letters and look-alike characters are all caught.
Anywhere young people write, safeguarding duties apply: learning platforms, homework submissions, club sign-ups, intranet chat. Staff need to catch bullying language and profanity early without reading every message.
Validate portal forms and messages as they are sent, scan uploaded documents, and let teachers check coursework in Word with a single click on the ribbon.
Patient feedback forms, complaints and care notes are part of a permanent record and may be read by families, regulators or courts. Language that is offensive, even when quoting a distressed patient, needs to be identified and handled consistently.
Flag offensive content in feedback as it arrives so it can be routed appropriately, and audit historical records in the database so nothing surprising surfaces during an inspection or disclosure.
HR systems accumulate free text from many hands: applications, referee comments, appraisal notes, grievance records, intranet posts. Offensive language in these records exposes the organisation in tribunals and undermines dignity-at-work policies.
Scan uploaded CVs and reference documents, validate intranet posts and appraisal forms at submission, and run the database scan across the HR system before annual audits or a system migration.
Communities live or die by their tone. Toxic chat drives members away and attracts platform penalties, while players work hard to sneak slurs past simple filters with symbols, spacing and reversed spelling.
Wire the text endpoint into chat, nickname and profile-bio creation. Fuzzy, reverse, leet and homoglyph matching catch disguised words in under a millisecond per call, and the decision rules can block, redact or just warn.
A single unfortunate word in a client deck, a print run or a social caption is expensive and very public. Copy passes through many hands and tools, and the final check is usually a tired pair of eyes late on a deadline.
Scan the PowerPoint deck, the Word manuscript and the Excel content calendar from the Metis Halo ribbon tab, and scan caption feeds or CMS entries through the API as part of the approval workflow.
Councils, regulators and law firms publish consultation responses, release documents under freedom-of-information rules and keep case files that will be scrutinised. Redaction must be consistent and defensible, not ad hoc.
Batch-scan document sets before release, clean consultation responses before they are published, and keep the job reports as evidence of what was checked and when.
Legacy systems hold free text nobody has looked at in years: old notes, imported comments, product descriptions written by suppliers. Migrating it wholesale moves the problem into a shiny new system, often one that is more visible to customers.
Point the database scan at the tables and columns to be migrated, or export them and run a file scan, then use the reports to fix or clean records before the cut-over.
Support conversations are emotional on both sides. Offensive language from customers needs handling with care, and a heated reply from an agent can become a screenshot on social media. Knowledge-base articles are edited by many people and are indexed by search engines.
Scan incoming tickets to route abusive ones to a senior agent, check outgoing replies before they send, and validate knowledge-base articles when they are saved.
Any site that lets the public write is a target for abuse, spam and hate speech, and platforms are increasingly held responsible for what they host. Word-list filters are trivially bypassed; human moderation does not scale.
Validate comments, bios and listings at submission with the form scan, scan uploaded text and CSV files, and sweep existing content tables periodically with the database scan so older posts meet the same standard.
| Base URL | https://api.metishalo.com |
| Authentication | X-Api-Key header with your personal key (see My Account) |
| Cost per scan | Text / form 1, file 1, database 3 credits; Office add-ins 1 credit per 1,000 scans |
- Exact token matching
- Fuzzy matching (Levenshtein distance)
- Reverse word detection
- Obfuscation detection (zero-width chars, spacing)
- Category-based rules & decisions
- Text cleaning / redaction