An examiner or auditor reviewing an adverse media check wants to see the search itself: which spellings of the name, which words, which languages, which years, and why each result was set aside. A file that says "no adverse media found" answers none of that. A check on one name is the smallest unit of adverse media screening, which runs the same kind of search across a whole book of customers. The steps here cover one name. Each adds a line to the record.
When do you need an adverse media check?
FATF's examples of enhanced due diligence (EDD) for higher risk relationships include more information on the customer, such as "information available through public databases, internet" (interpretive note to Recommendation 10, paragraph 20, June 2026 text). UK regulation 33(5) names additional independent, reliable sources and a better understanding of the customer's background.
The EU's AMLR lists additional information on the customer and beneficial owners among its EDD measures (Article 34(4)(a)). It applies from 10 July 2027. US banks must understand each customer relationship well enough to build a customer risk profile (31 CFR 1020.210(a)(2)(v)(A)).
Triggers include a high risk rating, a PEP match, unusual activity and a periodic review. A new majority owner at a corporate customer is one such event. Search the owner before the change is approved.
What changed in 2026
- 30 June 2026. S.I. 2026/621 rewrote UK regulation 33(1)(b), so its country trigger for EDD is now a "FATF call for action country"; other higher risk cases sit under 33(1)(g).
- 30 September 2026. AMLA, the EU's AML authority, published final draft standards on customer due diligence. Draft Article 21(b) lists information "to assess the reputation of the customer and the beneficial owners" among EDD inputs, and Article 23(g) names "reputable media publications" among the information that can show a customer's source of funds and source of wealth are lawful. Adoption is pending.
Can a Google search count as an adverse media check?
Picture two analysts (illustrative) who paste the same string into Google a week apart. Their first pages differ: one shows a court report the other never sees, and neither saved what came up. The Wolfsberg Group's Negative News Screening FAQs (2022) explain the gap.
Wolfsberg calls search engines content browsing tools rather than databases, built without financial crime risk in mind, so the most relevant item can sit far down the list (question 18). Website owners can purge, alter or archive content at any time, so one string can return different results on different days, and results also vary from user to user (question 20). In many countries the names of people under prosecution are not made public, and local data protection rules can allow unfavorable items to be removed.
So Google can carry an adverse media check for one name on two conditions: the string is written down before it runs, and every page you rely on is saved with its date. Skip either and nothing proves it ran.
How do you run an adverse media check, step by step?
Run the steps for each name. For a company, that means its legal and trade names and each director and beneficial owner.
- Collect the identifiers. Full name, date of birth or of incorporation, city, employer and role. They rule namesakes out later.
- Write the name block. Put each spelling in quotes for an exact match and join the spellings with OR in capitals; Microsoft's Bing search help says "You must capitalize the NOT and OR operators". Wolfsberg's transcription example gives Piotr Czajkowski, Pyotr Tchaikovsky and Piotr Chaykovskiy as forms of one name (question 11).
- Add one risk category at a time from Table 1, so a string reads ("Jane Doe" OR "Doe Jane") (bribe OR bribery OR kickback). Stage words such as alleged or probe swell the results, so Wolfsberg suggests precise words for the wrongdoing, such as bribe or fraud (question 19).
- Translate the keywords for each country where the person lives or trades, and narrow to local outlets with site: and the News filter (Google Search Help).
- Set the period with after: and before:, for example after:2016/01/01. Policy sets the lookback. Wolfsberg says a multi year lookback may fit, longer or shorter by type of offense, and that after the first review only new items may need screening (question 10).
- Run the name through the free adverse media check from complead, formerly Sanction Scanner. It scans global news outlets in multiple languages for financial crime coverage, regulatory actions and court records, matches every alias and spelling, and shows a story carried by many outlets once. The PDF report goes to your work email; no account is needed.
- Save every page you rely on, with its address and the date you read it.
Table 1. Search strings by risk category (a starting set; translate the keywords for each market)
| Risk category | Keywords to pair with the name block | What a relevant hit looks like |
|---|---|---|
| Money laundering | "money laundering" OR launder OR "proceeds of crime" | The person is a suspect, defendant or convict, or owns a firm used to move the funds |
| Bribery and corruption | bribe OR bribery OR kickback OR corruption OR embezzlement | Payments to officials, procurement kickbacks or misuse of public money by the person or their company |
| Fraud | fraud OR scam OR Ponzi OR forgery | The person ran or knowingly joined a scheme, rather than losing money to one |
| Tax crime | "tax evasion" OR "tax fraud" OR "VAT fraud" | A criminal case or penalty for evaded tax, not a routine tax appeal |
| Terrorism and terrorist financing | terrorism OR "terrorist financing" OR extremist | Funding, supplying or belonging to a designated or proscribed group |
| Sanctions evasion | "sanctions evasion" OR "sanctions violation" OR "export control" | Trade or payments routed around sanctions, or an enforcement notice naming the person |
| Organized crime and trafficking | "organized crime" OR trafficking OR smuggling OR cartel | Membership of, or services to, a criminal network, including drug, human and wildlife trafficking |
| Regulatory action | fined OR penalty OR "license revoked" OR disqualified | A supervisor's fine, a withdrawn license or a director ban, confirmed on the regulator's own site |
Most categories come from the Wolfsberg FAQs (question 5); fraud, sanctions evasion and regulatory action extend that list.
How do you tell whether a hit matters?
Ask four questions of each hit, in order, and write down each answer.
Is it the same person? Wolfsberg's grounds for discounting a false positive include a clear difference in the name or in a key identifier such as date of birth or date of incorporation; profession and residence are weaker criteria (question 24). Its sample rule closes a hit when the published birth date differs by a year or more (question 22).
Is the source credible? Editorial oversight, named dates, places and figures, and the same facts in another reputable outlet add weight; opinion pieces, social media posts and pages anyone can edit take it away (question 14). AMLA's draft Article 8 adds whether a source is official, current and accurate.
What stage, and what role? Wolfsberg sorts a matter by stage: allegation, investigation, charge, conviction (question 5). An acquittal is news too. A customer named as plaintiff is not the subject of the story (question 24), and the same goes for a victim or a witness.
Is it material? A speeding fine falls outside financial crime (question 6); a bribery report inside your lookback does not.
Allegations can turn out false. In Wolfsberg's own example, EDD on a new corporate customer satisfies the firm that allegations against the owners are false (question 7).
What does a worked adverse media check look like?
Take a UK distributor (illustrative) that is adding a Romanian packaging supplier, Ambalaj Nord SRL, whose director is Ion Popescu, born in 1967. Both names are invented, and the director's was chosen to look like a common Romanian name, so namesakes are the first problem. The know your supplier checks have already confirmed the company and its owners.
Run 1, in English. The name block with the bribery and fraud words returns about 60 results. Nearly all concern other men of the same name, and their towns and ages rule them out.
Run 2, in Romanian. The name block, "Ambalaj Nord" and mită OR "luare de mită" OR "dare de mită" return three results. One is a 2019 regional newspaper report that prosecutors were investigating the firm's director over a bribe to a city hall official for a packaging contract; the other two repeat it. Town, company and stated age match, so this is the same man, reported by an edited regional paper, at the investigation stage.
Run 3, the follow up. The same names with "trimis în judecată" (sent to trial), condamnat (convicted), achitat (acquitted) and clasat (case closed), after:2019/06/01, return nothing.
An empty result after 2019 says nothing about how the case ended. The analyst escalates: a written account from the director, a request for the court record, and an approver's decision on file.
What should the file record about the search?
Write the record as you search. Each line below answers a question that an auditor in the UK or EU, or an examiner in the US, will ask.
- Who searched and when, so the check can be dated against later news.
- Tools and engines used, including the free check.
- The exact strings and their languages, so anyone can rerun them later with the same settings.
- The period and settings: date range, filters, results language and any region setting.
- Results read per run.
- Each hit kept or dismissed, with the reason. Wolfsberg asks for the rationale on false positives as well as true matches (question 24).
- Captures of the pages you relied on.
- The decision, who made it and the next review date.
Where does a one off check stop and screening start?
Eight months later (still illustrative), a national paper reports that the supplier's director has been sent to trial. Nobody at the distributor searches again, because the check was a one off.
The free check has the same limit. It answers for the day it runs, and sanctions and PEP lists need checks of their own. A book of customers or suppliers needs screening that keeps running. In complead's adverse media screening and monitoring, NLP classifies negative news into risk categories, and API triggers start or stop monitoring for each name. The platform logs every search, including those with no match, with the searcher, the terms, the time and the result. A person still decides each hit, and the record shows who did.
The next question is what a material finding sets in motion. FATF's examples point to enhanced due diligence: more information on the customer, senior management approval and closer monitoring of the relationship. File that decision beside the search that started it.
