Skills-First Career Strategy & Cross-Border Digital Authority

The market has not replaced pedigree with proven capability. It has added a skills filter on top of pedigree, and most of the people selling you a “skills-first portfolio” are describing the filter, not the hiring decision.

Harvard Business School and the Burning Glass Institute, in their 2024 report Skills-Based Hiring: The Long Road from Pronouncements to Practice, measured what happened after firms publicly dropped degree requirements. Jeanne Meister, writing in Forbes on 6 January 2026, restated the finding that still governs this subject: 85% of companies report adopting skills-based hiring, and 0.14% of hires sit in roles where those degree requirements have actually been eliminated. That is not a rounding error. For every thousand hires after the policy change, roughly one person is different.

I would treat that 0.14% as the starting constraint, not as a footnote. You still need a skills-first architecture, because LinkedIn Recruiter, EURES, ESCO-mapped public employment services, and (from late 2027) the EU Talent Pool now search on skills taxonomies rather than job titles. You should not confuse being found with being hired. The rest of this piece is about making capability machine-readable across those systems, then producing the evidence a sceptical hiring manager, a visa officer, and a search index will actually accept.

The posting changed. The screen did not

Burning Glass and HBS split the firms that announced skills-based hiring into three groups. About 37% genuinely changed selection and hired more people without degrees. About 45% were “in name only”: the posting lost the bachelor’s line, the shortlist did not. About 18% tried and walked it back. Add the last two and you get 63% of self-described adopters whose behaviour is unchanged or reversed.

Indeed’s Hiring Lab, looking at November 2025 postings, found the raw share requiring a bachelor’s degree had risen to 19.3%, from 16.6% in 2023. After adjusting for occupational mix, the degree-adjusted share sat around 18.6%. The 2019-to-2023 drop in degree language on job boards was real. The 2023-to-2025 rebound is also real. Anyone still writing as if “the degree is dead” is working from the first half of that series.

LinkedIn’s own Future of Recruiting 2025 report (survey of 1,271 talent professionals in September 2024, plus platform data from more than a billion members) is more useful than the slogans built on it. Companies whose recruiters run a disproportionate share of searches through the skills filter are 12% more likely to make what LinkedIn calls a quality hire (equal weight on inbound demand, one-year retention, and internal mobility). Over nine in ten TA professionals say accurate skills assessment matters for quality of hire. In 2023, 26% of paid job posts on LinkedIn did not require a degree, against 22% in 2020.

Read that carefully. A 12% lift in a composite quality metric among firms that already search by skills is not “pedigree is over.” It is “if the recruiter actually uses the skills filter, the shortlist is slightly better.” Your job is to be in that filter. Your second job is to survive the human screen that still, in most firms, wants a recognisable school, a recognisable employer, or a recognisable visa path.

Michele Tuccio, Shizuka Kato and Roland Tusz, writing on the OECD Skills and Work blog on 21 May 2026 ahead of the 18 June launch of A Skills-First Labour Market, defined the policy shift in one sentence I will keep using:

skills-first approaches – approaches that focus on what people can do rather than on how or where they acquired their abilities

That is the official definition. It is also incomplete for a person who needs to cross a border. Immigration statutes still ask how and where. The revised EU Blue Card Directive lets some ICT specialists substitute three years of relevant experience in the last seven for a degree; Germany’s 2026 Blue Card still prices the role at EUR 50,700 gross a year for the standard threshold and EUR 45,934 for shortage occupations (STEM/IT), figures in force from 1 January 2026. Spain’s Orden PJC/44/2026, in force 30 January 2026, set the general Blue Card floor at EUR 39,269.92 (1.4 times the INE average wage) and a reduced floor of EUR 31,415.94. Skills open the door of the matching engine. Salary, contract length, and a qualification or a documented experience substitute still open the visa.

Skills are a taxonomy problem before they are a branding problem

The OECD report launched on 18 June 2026 (webinar with Manuela Geleng of DG EMPL, Soon Joo Gog of SUSS-IAL, Yustina Saleh of Burning Glass, Mark Pearson, Glenda Quintini, and Michele Tuccio) treats a common skills language as infrastructure, not as copywriting. The OECD/SUSS-IAL Skills-First Readiness and Adoption Index, published October 2025, scores 30 countries on three domains: a skills-first learning ecosystem, talent recognition, and an enabling environment. Twenty-five indicators. Higher score means the country’s employers, public employment services, and education system already talk to each other in skills, not titles.

If you live in a high-index country, the public systems will parse you. If you do not, you have to do the mapping yourself.

The European classification that now sits under EURES, Europass, and the forthcoming EU Talent Pool is ESCO. Version 1.2.1 shipped in December 2025. DG EMPL’s own note is dry and useful: a minor update, no concepts added or removed, translations cleaned, duplicate preferred/alternate labels stripped across 28 languages, occupation–skill links trimmed where an occupation had accumulated an absurd skill list, green labels revised. The database is downloadable as CSV, JSON-LD, ODS, RDF, TTL and XML, and there is an API. Each skill has a URI. That URI is the thing you want in a machine-readable profile, not a paraphrase you invented.

ESCO is not LinkedIn’s skills graph. LinkedIn’s graph is proprietary, English-heavy, and biased toward the job titles its members already use. O*NET is the US Department of Labor’s taxonomy and still sits behind a lot of US federal and contractor hiring. A skills-first portfolio that only speaks LinkedIn will miss EURES and the Talent Pool. A portfolio that only speaks ESCO will miss Recruiter Boolean. I would maintain a three-column map: your working name for the skill, the ESCO preferred term plus URI, the LinkedIn skill label as it appears in Recruiter filters. Update it when ESCO or LinkedIn changes the label. Do not update it because a coach told you to “sound more senior.”

World Economic Forum Future of Jobs Report 2025 (survey of more than 1,000 employers covering about 14 million workers in 55 economies) still has analytical thinking as the most cited core skill, named by 69% of employers, then resilience/flexibility/agility at 67%, leadership and social influence at 61%. The fastest-growing skills to 2030 are AI and big data, networks and cybersecurity, and technological literacy. Skill instability (the share of a worker’s skill set expected to be transformed or outdated between 2025 and 2030) is 39%, down from 44% in the 2023 edition and 57% in 2020. WEF’s own reading is that more people have already trained: 50% of workers had completed some upskilling or reskilling, against 41% in 2023.

Those WEF lists are demand signals. They are not your skill list. Copying “analytical thinking” into LinkedIn because 69% of surveyed employers ticked it is how you become unsearchable. Recruiter filters on exact or near-exact skill strings. “Analytical thinking” is a transversal competency in ESCO and a noise word in most Boolean strings. “PyTorch,” “SOC 2 Type II audit,” “IFRS 17,” “German C1,” “ISO 27001 lead implementer” are strings a human being will type.

What LinkedIn Recruiter actually queries

LinkedIn does not publish Recruiter ranking weights. Treat any blog that gives you “skills 35%, activity 25%, SSI 20%, geo 20%” as fiction until LinkedIn’s Economic Graph team puts it in a methods note. What LinkedIn does publish is more limited and more reliable.

The Talent Blog’s long-standing claim, still repeated in 2026 product copy, is that members who list at least one skill get up to twice as many profile views and connection requests and up to four times more recruiter messages, and that nearly half of hirers on the platform use skills data to fill roles. “Up to” is a ceiling. It is still the official number.

Future of Recruiting 2025 is explicit about how they defined “skills-based search” in the quality-of-hire analysis: companies where 25% or more of recruiters sit in the top quartile for searches that use the skills filter. The filter exists. A large share of recruiters still do not use it as their primary move. They search title, location, current company, then skills.

So the profile that wins the filter has:

The headline and current title containing the occupation string the recruiter will type (ESCO occupation or the local job title, not a creative internal title). “Staff Engineer, payments risk” beats “Problem solver in fintech.”

A Skills section whose first 10–15 entries are the Recruiter filter strings for the roles you want, in the language of the market you want, spelled the way LinkedIn stores them. LinkedIn lets you reorder. Order is a ranking input for how the profile renders; it is also how you stop “Microsoft Office” sitting above the skill that would have got you the InMail.

Experience bullets that repeat those skill strings in context, with an artefact. Recruiter’s keyword search reads the whole profile. A skill that appears only in the Skills section and nowhere in Experience is a weaker match than one that appears in both.

Location and work-authorisation fields filled in as the destination country understands them, not as your home HRIS labelled you. “Open to work” without a country, a work-permit status, or a remote/hybrid preference dumps you into the geo bucket the algorithm already over-weights against cross-border candidates.

Activity in the last 30 days is real as a distribution signal (dormant profiles are deprioritised in several LinkedIn surfaces). It is not a substitute for the skill match. I would not spend a working week manufacturing comments so an unpublished weighting can maybe move. I would spend that week putting one public artefact behind each of the five skills you want to be hired for.

Ryan Roslansky told the market in 2025 that job postings requiring AI literacy had risen more than 70% year on year. That is a demand spike, not a search syntax. If you add “AI” as a skill with no tool, no domain, and no artefact, you have joined a very large, very noisy set. If you add “evaluation harnesses for RAG systems in regulated insurance” and link a repo or a paper, you have given both the filter and the human something to do.

The portfolio is evidence, not a PDF

A skills-first portfolio is a small set of artefacts a stranger can inspect without you in the room. It is not a Notion page of adjectives. It is not a “personal brand.” Google’s own documentation is the right standard here, because the same stranger will often find you through search or an AI overview before they open Recruiter.

Google Search Central, in Creating Helpful, Reliable, People-First Content, says:

After identifying relevant content, our systems aim to prioritize those that seem most helpful. To do this, they identify a mix of factors that can help determine which content demonstrates aspects of experience, expertise, authoritativeness, and trustworthiness, or what we call E-E-A-T.

And then the sentence most career coaches skip:

While E-E-A-T itself isn’t a specific ranking factor, using a mix of factors that can identify content with good E-E-A-T is useful.

Danny Sullivan, when he was Google Search Liaison, was blunter: there is no E-E-A-T score. John Mueller in 2025: you cannot “add E-E-A-T” to a page. The Search Quality Rater Guidelines (the 11 September 2025 PDF that raters still use) put Trust at the centre of the family:

Trust is the most important member of the E-E-A-T family because untrustworthy pages have low E-E-A-T no matter how Experienced, Expert, or Authoritative they may seem.

That is the digital-authority problem in one clause. You do not become a subject-matter authority by declaring it on LinkedIn. You become one when third parties (journals, conference programmes, standards bodies, named employers, public issue trackers, court-admissible credentials) repeat a specific claim about what you can do, and when the artefact behind that claim still exists when a rater, a recruiter, or a visa officer clicks it.

I would build the portfolio as four layers, in this order, because the later layers are worthless if the first one is missing.

Layer 1: a skill inventory that a machine can join

One row per skill. Columns: working name, ESCO preferred term and URI, LinkedIn label, O*NET or local public-employment equivalent if you target the US or a specific PES, evidence URL, date last demonstrated, language of the evidence, whether it is a regulated profession (yes/no). Cap the public list at the skills you can defend in a 20-minute screen. A 50-skill dump is how ATS parsers and Recruiter both decide you are stuffing.

If the skill is language, use the CEFR label the destination uses (C1 German, B2 French), not “fluent.” If the skill is a regulated profession (nurse, architect, lawyer, some engineers), the inventory is not the portfolio. The competent authority’s recognition decision is. Skills-first does not waive that.

Layer 2: artefacts that show the work, not the story

For software: a public repository with a pinned version, a README that states what the system does, the failing case, and the constraint (latency, regulation, data residency). A gist of an error you diagnosed is more convincing than a “built scalable systems” bullet. For data and ML: an evaluation set, a metrics table, a note on leakage. For design: the constraint and the discarded option, not a Dribbble moodboard. For operations, tax, compliance, clinical, teaching: a redacted work sample, a published protocol, a named contribution to a standard, a teaching evaluation with a date.

Micro-credentials belong here only if they are verifiable. Europass Digital Credentials, sealed under eIDAS, mapped to ESCO and parked in a Europass wallet, are the European form that public systems can check. A PDF badge from a MOOC platform with no issuer public key is a decoration. The OECD report treats modular and micro-credential learning as part of the skills-first architecture; it does not treat every badge as recognition of prior learning. Recognition of prior learning is a legal process in many OECD countries. Do not confuse it with completing a four-hour course.

Layer 3: a canonical URL that search and LLMs can cite

One domain you control. Real name in the byline. Dates on every piece. An about page that states who you are, where you are based, and how to verify you (ORCID, company page, GitHub, court or chamber registration if relevant). Original work on a narrow subject, maintained. Google’s helpful-content guidance asks who created the content, how, and why. Answer those on the page. Do not answer them with a slogan.

Authority that “transcends geographical borders” is mostly this: a page that a recruiter in Munich, a hiring manager in Singapore, and an EU Talent Pool matching service can all retrieve, in a language they read, with an artefact they can open. Translation is part of the work. A German-language technical note on the same repo you already documented in English is worth more than a fourth LinkedIn post in English.

Layer 4: third-party repetition

Citations, conference talks with a programme page that still resolves, issues you filed that maintainers merged, standards comments with your name on the submission, journalism that quotes you on a fact rather than a vibe. This is slow. It is also the only layer Recruiter cannot fake for you and the only layer Google’s raters treat as reputation rather than self-description.

The EU Talent Pool will match on skills. It will not waive the permit

Regulation (EU) 2026/1047, signed 29 April 2026, published in the Official Journal on 12 May 2026, entered into force on 1 June 2026. The Commission’s Migration and Home Affairs page, updated around 10 June 2026, is clear on timing and scope. The platform is the first EU-wide tool for recruiting jobseekers who reside outside the Union into shortage occupations in participating Member States. Member State participation is voluntary. The Commission will now build the platform. Full operation is expected by the end of 2027.

Jobseekers register with skills, qualifications, work experience and language knowledge. Employers in participating states publish vacancies. Matching is the point. Talent Partnerships (including the EU–India gateway-office pilot) get a flag on the profile. None of that is a work permit. The permit remains national: Blue Card, national skilled-worker permit, seasonal, ICT, whatever the Member State runs. A beautiful ESCO-mapped Talent Pool profile that does not meet the Blue Card salary floor, or that sits in a non-participating Member State’s shortage list, is a matching success and an immigration failure.

If you are already inside the EU, EURES and national PES systems already speak ESCO. Use them. The Talent Pool is for people outside. If you are a US, UK, Indian, Brazilian, Nigerian, or Philippine national targeting the EU from abroad, the 2026–2027 window is for building the profile in the taxonomy the platform will ingest, not for waiting until the UI exists.

I would not treat the UK Skilled Worker route, US H-1B, or Canadian Express Entry as “skills-first” in the OECD sense. They are occupation-list and points systems with their own dictionaries (SOC 2020 in the UK, O*NET-adjacent codes in parts of US and Canadian processing). Map your inventory into those codes separately. Do not assume an ESCO URI impresses UKVI.

How talent algorithms fail you, ranked

Wrong occupation string. The recruiter searches “Wirtschaftsprüfer,” you wrote “assurance professional.” Fatal in German-language Recruiter and in most DACH ATS. Fixable in an afternoon if you swallow the local title.

Skills listed in the wrong language for the market. English skills on a French public-employment profile. Partial match at best.

Location set to a country where you cannot work. Recruiter geo filters are blunt. “Open to relocate” is weaker than a target city plus a one-line work-authorisation status.

Degree still used as a hard ATS filter even after the posting dropped it. This is the 45% “in name only” group. You cannot fix their ATS. You can put the degree-equivalent experience in the first two lines so a parser that still looks for “Bachelor” does not drop you before a human who might have meant the policy.

AI-written profiles that read like every other AI-written profile. LinkedIn’s 2025 report put GenAI use in recruiting at 37% of organisations actively integrating or experimenting, up from 27% a year earlier, with about 20% of the work week saved among those users. Hiring Assistant and related 2026 Recruiter features infer skills from your profile text. Homogeneous prose makes you indistinguishable. Specific artefacts make you expensive to confuse with the median.

Regulated-profession blindness. A skills-first medical or legal portfolio that skips the host-country recognition file will lose at the permit stage even if Recruiter loves you.

Time zones and overlap hours left blank. Cross-border remote hiring still dies in the “can they be in standup” conversation. Put overlap hours on the profile. It is a skill in the WEF sense (resilience, flexibility) and a logistical fact.

What I would do in the next 48 hours, given a full-time job

Pick one destination market and one occupation, not “global talent.” Write the ESCO occupation and the local job title on a line.

Export your current LinkedIn skills. Delete any you cannot evidence. Reorder the rest to the Recruiter strings for that occupation. Add the five you are missing if you can evidence them this month.

Put one public artefact behind each of those five. If you cannot publish the work, publish a redacted method, a talk, or a credential the issuer will still verify in 2027.

Register or update Europass if the EU is in scope; map the same five skills to ESCO URIs. If the US is in scope, check the O*NET task list for that occupation and make sure your bullets hit the tasks, not the marketing synonyms.

Read the host-country permit page for that occupation (Blue Card salary table, Skilled Worker going rate, or the national equivalent) dated 2026. If your current pay cannot meet it, the skills strategy is a mobility strategy for a later date, or a remote-from-home strategy. Do not pretend the algorithm will fix a salary floor written into an order in the BOE or the Bundesgesetzblatt.

The mistake that costs the most is performing skills-first for a labour market that has not implemented it, while neglecting the two systems that have: the matching engines that already filter on taxonomies, and the immigration rules that still price and credential you like 2019. Build for those. Let the press release about the death of the degree wait until Burning Glass prints a number that is not 0.14.