All In
FR EN
Back to blog
allinmedia 18 min read

LinkedIn's conversational AI search: how to prospect in plain English

| By Patrick de Carvalho

Contents

I have watched LinkedIn's search bar change shape for more than twenty years, since I joined the platform in April 2004. For most of that history, finding the right person meant learning its syntax: quotation marks, AND, OR, NOT, a dozen filters stacked in the right order. On November 13, 2025, LinkedIn's product team removed that requirement for Premium subscribers in the United States. LinkedIn's conversational AI search is a natural-language people-search tool: you type a sentence the way you'd describe a prospect to a colleague, and the system returns a filtered list of matching profiles, drawn from LinkedIn's professional graph, without you touching a single dropdown. This piece covers how it works, what it costs next to Sales Navigator, a framework for writing queries that actually return something useful, and the parts of the rollout nobody at LinkedIn has explained yet.

In short: LinkedIn's conversational AI search, live for US Premium subscribers since November 13, 2025, lets you describe a prospect in a full sentence instead of stacking filters. It understands job-title synonyms and relationship context that Boolean search cannot express, but it has no CRM export, no alerts, and no confirmed availability outside the US as of this writing. Sales Navigator still wins on account management; conversational search wins on discovery.

What actually shipped, and when

LinkedIn's own announcement is worth reading precisely, because most coverage since has rounded off its edges. The feature went live for Premium subscribers in the US, described in LinkedIn's own headline as exactly that: a new AI-powered people-search experience for Premium subscribers in the US. Four people from the product organization discussed the launch publicly: Tomer Cohen, LinkedIn's Chief Product Officer at the time, Rohan Rajiv, Senior Director of Product, Gyanda Sachdeva, VP of Product Management, and Wenjing Zhang, VP of Engineering.

Cohen's framing of the feature is the clearest single sentence LinkedIn has published on it: "Describe what you need in your own words, and we'll surface the people who can help." The examples LinkedIn itself used to demonstrate the tool tell you more than the marketing copy does. One became the reference case cited in nearly every write-up since: "ex-coworkers who became founders in healthcare in NY." A single sentence, four criteria (a past working relationship, a career transition, an industry, a metro area), resolved without a single filter click. LinkedIn's team offered others in the same announcement: "angels with FDA experience for an early biotech," and "Spanish-speaking school counselors in Austin focused on learning differences." None of those three queries could be built with LinkedIn's classic filters. There is no filter for "used to work with," none for "has FDA experience," none for "speaks Spanish and focuses on learning differences." That gap is the entire point of the feature.

Under the hood, the system runs on LinkedIn's professional graph: hundreds of millions of members, their job history, skills, education, posts, and connections, reprocessed by generative models that decompose a sentence into the criteria a classic filter search would have needed spelled out one by one.

Boolean search, the strings of AND, OR, and NOT that LinkedIn users have typed for two decades, works on exact matching. Conversational search works on inference. Three differences matter most for a sales team.

Semantic matching replaces literal matching. A Boolean search for "head of sales" returns exactly that string. It misses "VP Sales," "Director of Revenue," and "Chief Revenue Officer," unless you build the OR chain yourself and happen to guess every synonym in use. Conversational search understands that these titles cluster around the same function and returns all of them without being told to.

Relationship context becomes a searchable criterion. This is the structural change that matters most for prospecting. You can ask for people who worked at a named competitor and now work somewhere in your target category. Filter-based search has no field for "used to work at." It only knows where someone works today.

Multi-criteria complexity stops being capped. Sales Navigator's filter panel has a practical ceiling: you can stack filters, but each one narrows the same static list. Conversational search absorbs a five- or six-clause sentence and treats every clause as a simultaneous condition, which is closer to how a human recruiter or account executive actually thinks about a target list.

Conversational search vs Sales Navigator, side by side

Neither tool replaces the other. They sit at different stages of the same pipeline.

Criterion Sales Navigator Conversational AI search
How you query Predefined filters, Boolean operators Full sentences, natural language
Learning curve Moderate to steep Low, if you can describe the person
Job-title matching Exact or manually built OR chains Semantic, automatic
Relationship context Limited to 1st/2nd-degree connections Career history, "used to work at X"
Combinable criteria 20-plus filters, stacked sequentially Effectively unlimited, simultaneous
Lead lists and CRM sync Native export and CRM integration No native export
Alerts on job changes or posts Yes No
Verified email or phone No, on either tool No
Price (per LinkedIn's public tiers) $79.99 to $159.99 per month Included in LinkedIn Premium, $29.99 per month
Confirmed availability Global US Premium subscribers, at launch

The gap that stands out here is one Martal Group's own comparison of prospecting tools flags bluntly: Sales Navigator, for all its filters, still cannot hand you a verified email or a direct phone number, which is why teams layer on dedicated data providers such as Apollo or ZoomInfo when outreach depends on more than an InMail. Conversational search inherits that same gap. Neither tool was built to replace a contact database; both were built to help you find the right name faster.

The TRACE framework: five variables for a query that works

A vague sentence returns a vague list. After running enough queries through this kind of tool, five variables consistently separate a query that returns fifteen usable names from one that returns three thousand irrelevant ones. All In built this into a framework called TRACE, because that is functionally what you are doing: tracing a specific person through a graph of a billion-plus profiles, not filtering a static database.

T, for Title. State the function in plain language rather than a job-title string, and let the model handle the synonyms. "People who run procurement" will catch "Head of Procurement," "Supply Chain Director," and "VP of Sourcing" without you writing any of them down.

R, for Relationship. Name the connection you're after: former colleagues, second-degree connections, alumni of the same school, members of the same professional group. This is the criterion Boolean search cannot express at all.

A, for Account. Specify the industry and, where it matters, company size or funding stage. "At mid-market manufacturers" behaves differently from "at seed-stage manufacturing startups," and the model treats them as distinct filters.

C, for Coordinates. Geography still narrows a US B2B search meaningfully, whether that means a metro area, a region, or a specific pair of cities.

E, for Event. Add a time-bound signal: a recent job change, a funding round, a hire, a post on a specific topic. This is what turns a static list into a list of people who are, right now, more likely to answer.

A full TRACE query reads like this: "Second-degree connections who run procurement at mid-market manufacturers in the Midwest and changed jobs in the past three months." Precise, contextual, and not something any filter panel could assemble in one step. In the French original, All In's editorial team built the same five variables as CIBLE (Contexte, Intitulé, Business, Localisation, Événement); TRACE is the English scaffold on the same logic.

A dozen queries built for a US B2B pipeline

These are written the way a native English speaker would actually type them into the search bar, not translated from anything. They split into three jobs a sales or business development team typically needs done.

Finding new prospects

  1. "Founders of manufacturing companies in the Rust Belt who post about reshoring at least once a week"
  2. "VPs of Sales at Series B SaaS companies who changed jobs in the last six months"
  3. "Founders of fintech startups that raised a seed round and have between 10 and 50 employees"
  4. "Marketing directors at mid-size healthcare systems in Texas or Florida"

Mapping a target account

  1. "People who used to work at [competitor] and now hold leadership roles at [target company]"
  2. "Decision-makers at [target company] who are connected to someone in my network"
  3. "Alumni of [university] holding VP or C-suite titles in healthcare"

Watching a niche market

  1. "Cybersecurity experts based in the US publishing analysis on the SEC's new disclosure rules"
  2. "Climate tech investors who commented on a funding announcement this week"
  3. "Independent consultants specializing in applied AI for manufacturing"
  4. "HR leaders at construction companies in the Southeast who hired more than 20 people this year"
  5. "Board members at mid-market energy companies who are active on LinkedIn"

Swap in your own competitor names, target accounts, and regions. The structure is what transfers, not the wording.

Three ways sales teams are already using it

Chasing a funding signal. A fractional-CFO firm wants companies that just closed a round and are about to need financial infrastructure they don't have yet. The query: "Founders and COOs of Series A or Series B startups that closed a round in the last two quarters, based in the Bay Area, Austin, or Boston." Building this with filters would mean cross-referencing a funding database by hand, since Sales Navigator has no field for "recently funded." Conversational search resolves it in one step by reading career-history signals directly.

Mapping the real organization chart. An HR software vendor wants to know who actually influences a target account's technology decisions, not just who holds the VP HR title on the org chart. The query: "People at [target company] working on HR technology or HRIS projects who are connected to someone in my network." The relationship clause does the work here: it surfaces a warm path into the account that a title-only filter never would.

Watching a specialty market before it gets crowded. A lab-equipment manufacturer wants to reach growing biotechs before a competitor does. The query: "Directors of science and directors of manufacturing at biotech companies with 20 to 200 employees, actively hiring, in Cambridge, San Diego, or the Bay Area." The signal "actively hiring" is doing something specific: a company staffing up its lab team is a company about to need lab equipment, and that is the moment to be in front of them, not after.

What changed since the November 2025 launch

Two developments since launch matter enough to reshape how a team should think about this tool going into the second half of 2026, and neither made it into most of the coverage published in the weeks right after launch.

LinkedIn folded an AI Sales Assistant directly into Sales Navigator, aimed at cutting the administrative load of prospecting rather than the discovery step conversational search already handles. And LinkedIn opened an early beta of a Company Intelligence API, a data connection sales and marketing tools can plug into directly. The beta numbers, reported in February 2026 and self-reported by LinkedIn rather than independently audited, are striking enough to name with that caveat attached: a 287% increase in companies reached, 75% more marketing-qualified leads (prospects a marketing team judges ready to hand to sales), 96% more sales-qualified leads (prospects a sales rep judges worth a live conversation), and a 43% drop in acquisition cost.

Metric (early beta, Feb. 2026) Reported change
Companies reached +287%
Marketing-qualified leads +75%
Sales-qualified leads +96%
Customer acquisition cost -43%

LinkedIn also reworked how its feed ranks content, introducing what it calls a Depth Score: the algorithm now weighs how long someone actually engages with a post, whether they save it or send it privately, and whether a comment adds something, over a quick like or a generic "Great post!" That doesn't change how conversational search works, but it changes what shows up next to a prospect's name once you've found them, and it rewards exactly the kind of specific, sourced writing this piece is trying to model.

What it still doesn't do

None of this replaces a dedicated sales stack, and LinkedIn has never claimed it would.

No native CRM export. Sales Navigator saves lists and syncs them into Salesforce, HubSpot, or Pipedrive. Conversational search returns a browsable result set with no built-in export path. You copy names by hand or wait for a third-party tool to bridge the gap, and LinkedIn actively polices scraping extensions that try.

No alerts. Sales Navigator pings you when a saved lead changes jobs or posts. Conversational search is a one-time query, not a monitoring feed. Run it again next month if you want fresh results.

No verified contact data. Neither tool hands you an email address or a phone number. That gap is precisely what pushes teams toward Apollo, ZoomInfo, or similar providers once a name on a list needs to become an actual conversation, and reporting a 4-to-7x lift in meeting-conversion rates when that data layer is added on top of manual LinkedIn outreach, according to platform vendor data cited by Martal Group. Treat that figure as vendor-supplied, not independently audited, and weigh it accordingly.

Privacy settings still apply. The tool only surfaces public profiles or ones within your network's visibility settings. A prospect who has locked down their profile stays invisible no matter how the query is phrased.

Outside the United States: an honest status report

Here is where I'd rather tell you what isn't confirmed than guess. LinkedIn's own November 2025 materials are specific: the feature launched for Premium subscribers in the US, full stop, with no country list, no international date, and only the phrase "rolling out to all members soon" attached to it. Reporting since has said LinkedIn intends to add French, Spanish, German, and Portuguese language support during 2026, but I have not found a second, independent confirmation that this shipped, and LinkedIn has not published an updated timeline as of this writing.

That leaves UK-based Premium subscribers in a genuine gray zone. English is the feature's native language and the UK is one of LinkedIn's largest single markets, with roughly 44.6 million members, behind only the US at around 250 million among individual countries. Neither fact tells you whether a London-based account currently has access, because LinkedIn's rollout language names a country, not a language. If you're on a UK or EU account and want to know today, the only reliable test is checking your own search bar. Don't trust a screenshot from someone else's account as proof either way.

What to do this week, with or without access

Three moves are worth making regardless of whether your account has the feature yet.

Get your own profile ready to be found, by AI as much as by people. A widely cited March 2026 analysis from Axios, based on research from Profound, found LinkedIn to be one of the most-cited sources when AI chatbots such as ChatGPT and Perplexity answer professional questions. A profile with clear, specific language in the headline, About section, and experience entries gets picked up by that layer as reliably as it does by a human reader typing a conversational query. All In's own G.E.O. LinkedIn method, detailed on All In's methods page, walks through the specifics.

Write down your TRACE queries before you have access, not after. Five to ten queries covering your priority segments, drafted now, save you weeks of trial and error the day the feature reaches your account.

If you have Premium and an English-language interface, test it today. Access has been reported by some non-US Premium users with English set as their interface language, though this is not something LinkedIn has confirmed as a supported path, so treat any result as unofficial until your own account proves otherwise.

The instinct to write outreach messages that read like everyone else's applies here too, and it's the same failure mode covered in All In's piece on LinkedIn's "seems like AI slop" reporting button: a faster way to find a name doesn't fix a generic message once you've found it. Bring a number the prospect hasn't seen elsewhere, name the specific signal that made you reach out, and the tool that got you the name stops mattering to the person reading your message.

What I don't know

I don't know whether the French, Spanish, German, and Portuguese rollout that reporting described for 2026 has actually shipped by the time you're reading this. I don't know the feature's false-negative rate: how many good-fit prospects a query silently drops because the model read "biotech" or "fintech" more narrowly than you meant. I don't know whether the Company Intelligence API's beta numbers hold up outside the accounts LinkedIn chose to showcase; they are self-reported, from an early beta, and worth watching rather than trusting outright. And I don't know whether a UK or EU Premium account has real, supported access today. If you find out from your own account, that's more current than anything published here.

FAQ

Is LinkedIn's conversational AI search available outside the United States?

Not confirmed. LinkedIn's official launch, November 13, 2025, was for Premium subscribers in the US only, with no country list attached and no updated international timeline published since. Language support for French, Spanish, German, and Portuguese was reported as planned for 2026, but independent confirmation that it has shipped has not surfaced.

No. It's included in standard LinkedIn Premium, at $29.99 per month, not gated behind Sales Navigator's higher tiers ($79.99 to $159.99 per month). The two tools are complementary: Premium for natural-language discovery, Sales Navigator for list management, alerts, and CRM sync.

Does conversational search replace Boolean search operators?

It doesn't remove them, but it makes them optional for most searches. Boolean operators (AND, OR, NOT) still work in classic search and in Sales Navigator. Conversational search translates a plain-English sentence into that same underlying logic, enriched with semantic matching a manual Boolean string can't replicate.

Can I export results from conversational search into a CRM?

Not natively. Sales Navigator saves and syncs lead lists into Salesforce, HubSpot, or Pipedrive. Conversational search returns a browsable list with no built-in export, so you're copying names manually or waiting on a third-party bridge, while LinkedIn actively restricts scraping tools that try to automate that step.

Does it work as well in languages other than English?

The model was trained primarily on English-language data and performs best in it. LinkedIn has reported plans to extend support to French, Spanish, German, and Portuguese during 2026, though as of this writing that expansion has not been independently confirmed as live.

What data does the search actually use, and does it respect privacy settings?

It draws on public LinkedIn data: job titles, employers, education, skills, posts, and the connection graph. It only returns profiles that are public or visible under your account's network settings; a locked-down profile stays invisible regardless of how the query is phrased.

Sources

  1. LinkedIn News, November 13, 2025: official launch announcement, US-only availability, Tomer Cohen quote, example queries.
  2. Social Media Today, November 2025: launch corroboration, additional example queries.
  3. VentureBeat, 2025: technical architecture of LinkedIn's generative search models.
  4. Fortune, April 1, 2025: Tomer Cohen interview, product leadership context.
  5. Fast Company, 2025: reported plans for French, Spanish, German, and Portuguese language expansion.
  6. Rev-Empire, "LinkedIn's 2026 Updates: What B2B Sales Teams Need to Know": AI Sales Assistant, Company Intelligence API beta figures, Depth Score algorithm change.
  7. Affinco, "LinkedIn Statistics 2026": US and UK member counts, Premium subscriber figures, InMail response benchmarks.
  8. Martal Group, "LinkedIn Sales Navigator Alternatives for 2026": Sales Navigator data gaps, AI-investment figures, conversion-rate comparisons.
  9. Axios, March 10, 2026: LinkedIn's citation share in AI chatbot answers, Profound research.
  10. WebProNews, 2025: additional coverage of the conversational search launch.

All In: the query changed, the message still has to earn the reply

A better search bar gets you to the right name faster. It says nothing about what you write once you get there, and that's still the part nobody automates for you.

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.