LinkedIn profile optimization for executives, rebuilt for AI search
Contents
- What changed in 2026: LinkedIn search stopped matching keywords
- The seven pillars, in the order you should run them
- Positioning: write a headline a ranking model can use
- Rewrite: LinkedIn's AI assistant is a first draft, not a final one
- Outbound filters: what Sales Navigator queries is not what you wrote
- Featured: the only part of your profile you fully control
- Identity: the free badge most members still skip
- Link your calendar: from profile view to booked meeting
- Small screen: where a large share of your buyers read you
- Three hours, in sequence
- The numbers everyone quotes, and which ones survive a source check
- What I don't know
- The part of your profile no model can rank for you
- FAQ
- Sources
On February 7, 2026, Fedor Borisyuk and seventy-three co-authors posted a paper on arXiv titled "Semantic Search At LinkedIn," describing the machinery now between a buyer's question and your profile: embedding-based retrieval, a compact ranking model, a larger model acting as relevance judge. The line that matters sits in the abstract's first sentence: retrieval happens by meaning, not keyword overlap.
I have been on LinkedIn since April 2004, its first year, and profile advice has survived four platform redesigns essentially unchanged. It does not survive this one: packing your headline with pipes and job titles was a sound bet against a keyword index, a poor bet against a model trained to read for evidence. Below is the seven-part method I run, rebuilt on that mechanism, with every number traced to whoever published it.
In short: LinkedIn now ranks people search results with language models rather than keyword matching (arXiv 2602.07309, February 7, 2026), and conversational search reached all US members in April 2026. Evidence, not keyword density, is what your profile is judged on. The seven moves below take about three hours, six are free, each tied to a source you can open yourself.
What changed in 2026: LinkedIn search stopped matching keywords
Two things happened, eight months apart.
First, the engineering: profiles are retrieved by a bi-encoder that compares meaning, then reranked by a small language model distilled from larger teacher models, boosting ranking throughput by more than 75x under a fixed latency budget while holding quality close to the teacher model.
"Semantic search with large language models (LLMs) enables retrieval by meaning rather than keyword overlap, but scaling it requires major inference efficiency advances." Borisyuk et al., "Semantic Search At LinkedIn," arXiv 2602.07309, February 7, 2026
Second, the rollout: LinkedIn launched AI people search to US Premium subscribers in November 2025, then, per Social Media Today on April 15, 2026, expanded it to all US members: search now handles nicknames and misspellings, verification badges appear inline in results, and a short AI-generated summary explains why each profile surfaced. That is the whole argument: a model writes a one-line case for you, to a stranger, from whatever your profile gives it, and the profiles that win carry checkable facts, not a repeated title.
The paper covers AI Job Search and AI People Search only; Sales Navigator and recruiter search still run on structured fields (industry, headcount, seniority, years in role, geography), tracked on the All In blog.
The seven pillars, in the order you should run them
All In publishes this method as P.R.O.F.I.L.S., which in the original French spells the word for profiles: same seven jobs, same order, each step depending on the one before it.
| Letter | The move | Time | Free? |
|---|---|---|---|
| P | Headline and About rewritten for meaning-based retrieval | 45 min | Yes |
| R | AI assistant for a draft, own the final text | 30 min | Yes |
| O | Fix the structured fields Sales Navigator reads | 20 min | Yes |
| F | Rebuild the Featured section as a short funnel | 30 min | Yes |
| I | Get the verification badge | 5 min | Yes |
| L | Booking button, or the free workaround | 15 min | Button needs Premium |
| S | Check every section on a phone | 15 min | Yes |
Roughly three hours total, across a Tuesday afternoon, and six of the seven steps are free.
Positioning: write a headline a ranking model can use
Your headline runs to 220 characters, About to 2,600, per AuthoredUp's field limits, with desktop showing only the first 300 characters of About before the "See more" cut.
The advice that dominated the last decade, loading the headline with searchable terms separated by pipes, worked against lexical matching. Against a model comparing meaning, repeating "B2B sales" four times adds nothing: the second occurrence carries no new information for an embedding.
A specific claim with a specific object adds information. Weak, and I see this daily: "Passionate about digital transformation | Helping companies grow | Speaker", naming no buyer, no outcome, no market. Workable: "CEO at Northbound Freight | Cutting freight spend 12-18% for mid-market 3PLs | Logistics, US Midwest", naming role, outcome, buyer, and market.
Front-load the first 60 to 70 characters: that is what appears next to your name in search results and connection requests.
Rewrite: LinkedIn's AI assistant is a first draft, not a final one
LinkedIn ships a writing assistant that will produce your About section from your existing profile data. Use it, then delete most of what it gives you.
I say that as someone who publishes AI-assisted work and discloses it, a position I argued in the piece on LinkedIn's "AI slop" button: the problem is not the machine, it is that the draft reads like the platform average, exactly what you are trying to be distinguishable from.
The assistant gets you past a blank field with a structure to argue with. Three elements have to be yours, and no tool can supply them: your numbers (revenue grown, clients served, cost removed, with the period attached: "Grew ARR from $4M to $11M between 2023 and 2026" is checkable, "Proven track record of growth" is not), your position (something a competitor would argue with), and your named cases (industry, size, situation, result, anonymized if you must, but specific).
The test: could this sentence appear on a competitor's profile unchanged? If yes, it is doing no work for you.
Outbound filters: what Sales Navigator queries is not what you wrote
Sales Navigator filters on structured fields, not your prose: industry from a dropdown in your profile settings, not the word "manufacturing" in your About section; company size from your company page; seniority inferred from your current job title, so "Chief Storyteller" removes you from every filter for C-level executives; years in role from your start date; skills directly filterable. Twenty minutes of dropdown maintenance decides whether a buyer's filtered list contains you at all: your profile's company must match your company page, same industry, same location, or mismatches drop you from filtered results.
LinkedIn's talent blog states members who list at least one skill get up to 2x more profile views and connection requests, up to 4x more messages; you can list up to 50, pin five. It also reports 75% of hirers want context: a skill with no supporting evidence in your experience section is a claim without a case.
On the Social Selling Index, LinkedIn's Sales Solutions page says this:
"A high SSI score doesn't always represent the efficacy of a sales person or correlate with measurable sales outcomes." LinkedIn Sales Solutions, Social Selling Index page
That is the vendor telling you not to treat the score as an outcome: a free diagnostic on profile completeness, not a score to manage a team against.
Featured: the only part of your profile you fully control
The Featured section is the one block where you decide what a visitor sees. Most executives use it as a trophy shelf: a conference photo, a press mention, a two-year-old post. Treat it as four slots in sequence.
| Slot | What goes there | Job it does |
|---|---|---|
| 1 | A useful free resource with a real title | Capture a contact, or simply earn the next scroll |
| 2 | A case with numbers in it | Prove the claim your headline makes |
| 3 | Your signature piece of thinking | Establish that you have a position |
| 4 | Your booking link or contact page | Convert the visit |
Titles matter more than the assets: "My latest post" tells a visitor nothing, "How a 40-truck carrier cut deadhead miles by 19% in one quarter" tells them whether to click. Custom thumbnails beat automatic previews, and refresh the section quarterly: a newest item eighteen months old signals a dormant account.
Final Layer's February 2026 analysis puts the gain from a completed Featured section at up to 30% more profile views and connection requests, a vendor figure flagged in the source-check table further down.
Identity: the free badge most members still skip
LinkedIn crossed 100 million verified profiles in December 2025, out of roughly 1.3 billion members, call it 8%, free and five minutes.
LinkedIn's help center states verified members see 60% more profile views and 50% more engagement, current and published though its own number, and since April 2026 badges also appear directly in the result list.
Three routes exist, and the geography matters:
| Verification type | How | Where |
|---|---|---|
| Identity | CLEAR, with a government ID and a phone number registered locally | United States, Canada, Mexico |
| Identity | Persona, with an NFC-enabled passport | Many additional countries, including most of the EU and the UK |
| Workplace | Work email, Microsoft Entra Verified ID, or a company LinkedIn Learning or Recruiter license | Global, where the employer has enabled it |
The CLEAR route is North America only; outside it, use Persona. Workplace verification through your work email is available wherever your company has enabled it, an admin setting, not a purchase. Settings, then Verifications: LinkedIn walks you through it.
Link your calendar: from profile view to booked meeting
LinkedIn's standard path, read a post, open the profile, connect, wait, message, trade availability, loses people at every step, the waiting steps most of all.
LinkedIn Premium Business includes a custom call to action button that "appears on your profile, posts, messages, and in search results." Calendly and LinkedIn announced a direct integration on March 20, 2025, letting subscribers add a Book an Appointment button and schedule without leaving LinkedIn: more than 165,000 members already had a scheduling link at the time, and Premium subscribers using custom buttons average 25% more profile views. The most useful line:
"Over 90% of my sales meetings come from people who've never messaged me; they've been quietly following my content, and when they're ready, they book a time directly." Corrina Owens, founder and go-to-market advisor, quoted in Calendly's March 20, 2025 announcement
The booking button does not create demand: it removes the messaging step from demand that already exists.
Premium Business runs $69.99 per month in the US, or $539.88 annually. If that is not a spend you want this quarter, the free version is close: put the scheduling link in slot 4 of your Featured section and website field, named for what it does. "Book 20 minutes" beats "Website."
Small screen: where a large share of your buyers read you
The "70% of LinkedIn traffic is mobile" figure everyone repeats does not hold up: the last device split LinkedIn disclosed was 57% mobile, a decade ago (full picture in the source table further down). A majority of your visitors are on a phone, and nobody outside LinkedIn knows the exact share.
Open the LinkedIn app on your own phone, search your own name, and look at your profile the way a stranger would: About section cutoff, headline survival, Featured items visible before you swipe. Cutoff points shift with app versions, so measure your own profile rather than trust a number from an article, including this one.
Then fix in this order: the first line of your About section, since "See more" hides most of your text on a phone; the first sixty characters of your headline; banner text legibility at actual phone size, which kills most banners built in Canva at desktop scale; and the first Featured slot, possibly the only one a phone visitor ever sees.
Three hours, in sequence
Run the pillars in that order, since every later step depends on the claim you settle on in positioning: positioning (45 min), rewrite (30 min), outbound filters (20 min), featured (30 min), identity (5 min), link (15 min), small screen (15 min).
Then measure over the following month: profile views, search appearances, inbound connection quality, and how many people reach your booking link. LinkedIn gives you the first two natively; methodology in the All In methods, the weekly version in the All In newsletter.
The numbers everyone quotes, and which ones survive a source check
Every headline statistic in the genre, checked against its origin in August 2026.
| Claim in circulation | Who actually published it | When | Verdict |
|---|---|---|---|
| A photo gets 21x more profile views and 9x more connection requests | LinkedIn Sales Blog, Sean Callahan | Dec 28, 2018 | LinkedIn's own, unrefreshed in eight years |
| An All-Star profile gets 40x more opportunities | No primary LinkedIn page found | Unknown | Quoted everywhere, sourced nowhere I could reach |
| At least one skill listed: up to 2x profile views, 4x messages | LinkedIn Talent Blog | Current | Holds, far more modest than the folklore |
| Verified profile: 60% more views, 50% more engagement | LinkedIn Help Center, and the December 2025 milestone announcement | Dec 2025 | Current, first-party, vendor measuring its own feature |
| Featured section: up to 30% more views and requests | Final Layer blog | Feb 18, 2026 | Third-party, no sample or method published |
| SSI above 75 means 2.8x content performance | Not found | Unknown | LinkedIn now warns SSI may not correlate with sales outcomes |
| 70% of LinkedIn traffic is mobile | Third-party panels, no first-party figure since 57% at 414M members | Varies | Directionally right, precisely unknowable |
I published the 40x number myself, in the French version of this method, and in August 2026 could not find the LinkedIn page it came from. Until someone shows me the source, I am retiring it: an untraceable statistic is one someone else can use against you.
What I don't know
Five gaps, stated plainly.
I do not know whether the semantic ranking described in the LinkedIn paper applies to recruiter search or Sales Navigator. The paper names AI Job Search and AI People Search. I am assuming the structured-filter systems still behave as they always have, which is why pillar three exists, but that is inference on my part, not a documented fact.
I do not know how long profile edits take to propagate into search results, and neither does anyone quoting a number at you.
I do not know whether verification affects ranking, as opposed to display. LinkedIn reports more views for verified members and shows badges in results. Correlation between a badge and views is not evidence that the badge is a ranking input, and LinkedIn has not said it is.
I have no independent measurement of the Featured section claim. One vendor blog is not a study.
And I have no controlled test of the seven pillars against a holdout group. Nobody publishing this kind of method does, mine included. What I have is twenty-two years on the platform and the sources listed at the bottom, which is why every claim above is either attributed or flagged.
The part of your profile no model can rank for you
The retrieval system changed, and most of the advice written for the old one is now noise.
The strategic news is smaller and less comfortable: semantic retrieval rewards evidence, so the constraint moves off the copywriting and onto the underlying material. You cannot phrase your way into having a number, a named case, or a position that costs you something. With those, the seven pillars make them findable; without them, a well-optimized profile just gets you found faster by people who then find nothing. How I disclose my own working method, tools included, is on the All In about page.
A decent problem to have discovered on a Tuesday afternoon, three hours in, rather than in a first meeting.
FAQ
Does LinkedIn's AI people search change how I should write my headline?
Yes. LinkedIn's February 2026 paper describes retrieval by meaning, not keyword overlap, and conversational search reached all US members in April 2026. A specific claim naming an outcome, a buyer type, and a market beats a repeated keyword. Front-load the first 60 to 70 characters.
Is the LinkedIn verification badge worth it, and does it cost anything?
Free, about five minutes. LinkedIn's help center reports verified members average 60% more profile views and 50% more engagement, its own number. Since April 2026, badges also appear directly in search results. In the US, Canada, and Mexico, verification runs through CLEAR; elsewhere through Persona.
Do I need LinkedIn Premium to put a booking link on my profile?
Only for the native button. LinkedIn Premium Business, at $69.99 per month in the US, includes a call to action button on posts, messages, and search results, connected to Calendly since March 2025. Without Premium, use your Featured section and website field, labeled with the action.
How many skills should I list on my LinkedIn profile?
LinkedIn allows up to 50 skills, five pinned in your About section. Its talent blog reports that listing at least one brings up to 2x more profile views and connection requests, up to 4x more messages, and that 75% of hirers care about context, so a skill absent from your experience carries less weight than one your work history demonstrates.
Does the Social Selling Index still matter in 2026?
Less than the advice industry claims. LinkedIn's Sales Solutions page states a high SSI score does not always represent a salesperson's effectiveness or correlate with sales outcomes, and now steers users toward AI-based selling tools instead. Use the professional brand pillar as a free completeness check, not a score to manage a team against.
How long does it take for profile changes to show up in search?
LinkedIn does not publish a re-indexing interval, and any specific number you see quoted has no primary source behind it. What you can measure is your own search appearances, reported natively in your profile dashboard: change one thing, note the date, and watch that metric over the following weeks.
Is it safe to write my About section with LinkedIn's AI assistant?
Safe, and insufficient. The assistant drafts from platform-average phrasing, indistinguishable from competitors' text. Keep it for structure, then replace anything generic with your own figures, named cases, and a position someone could disagree with.
Sources
- Borisyuk et al., "Semantic Search At LinkedIn," arXiv 2602.07309, February 7, 2026: embedding-based retrieval, small language model ranker, LLM relevance judge, 75x throughput gain.
- Social Media Today, "LinkedIn expands AI-powered conversational search", April 15, 2026: rollout to all US members, AI-generated profile summaries in results, verification badges inline, flexible matching.
- Social Media Today, "LinkedIn Reaches 100M Verified Profiles", December 9, 2025: 100 million verified members, 60% more views, 50% more engagement, verified Pages figures.
- LinkedIn Help Center, "Verifications on your LinkedIn profile": identity, workplace, and education verification, CLEAR, Persona, DigiLocker, availability by country, no cost.
- LinkedIn Talent Blog, "How to Add Skills to Your LinkedIn Profile": up to 2x profile views and 4x messages, 50 skills maximum, 5 pinned, 75% of hirers want context.
- Sean Callahan, LinkedIn Sales Blog, "Picture Perfect", December 28, 2018: 21x more profile views, 9x more connection requests with a photo.
- LinkedIn Sales Solutions, Social Selling Index: SSI definition and the caveat on correlation with sales outcomes.
- LinkedIn Premium Business: $69.99 per month, $539.88 per year, custom call to action button, 15 InMail credits, 365 days of viewer data.
- Calendly newsroom, "Calendly Teams Up with LinkedIn", March 20, 2025: Book an Appointment button for Premium Business, 165,000 members with scheduling links, 25% more profile views with custom buttons, quotes from Corrina Owens, Liam Darmody, Naren Raghavan, and Ora Levit.
- Final Layer, "LinkedIn Featured Section: What to Add & How It Drives Leads", February 18, 2026: up to 30% more profile views and connection requests, no sample or method published.
- AuthoredUp, "LinkedIn Character Limits 2026": headline 220 characters, About 2,600 characters, About truncation around 300 characters on desktop.
- VentureBeat, "LinkedIn now has 414 million members, with 57% of traffic coming from mobile": the last first-party device split LinkedIn disclosed.
- Cognism, "100 Essential LinkedIn Statistics and Facts for 2026": approximately 1.3 billion registered members, United States as the largest national member base.
All In: a profile that gets found is only half the work
Semantic retrieval rewards evidence, which means the winning move is no longer a phrasing trick but a documented claim you can defend in a meeting. All In covers both halves, the mechanism and the material.
All In is the B2B media that decodes LinkedIn, expert blog, weekly podcast and newsletter for SME leaders and sales directors who want to turn LinkedIn into measurable growth. An original creation by Patrick de Carvalho, on LinkedIn since 2004. Motto: "I Never Lose."
Discover All In and get what matters on LinkedIn, every week.