Short answer
Write the search the way you would brief a colleague: name the role, the place and the two or three requirements that really matter, and leave nice-to-haves for a second round. In Essentiq the AI turns that description into filters (title and synonyms, location and radius, skills, education, companies), lists the candidates who meet all criteria first and explains each match, so you can check the interpretation and refine in plain language.
On this page
- How an AI candidate search reads your description
- The anatomy of a good search description
- Worked examples: weak vs strong descriptions
- Check the interpretation before you trust the result
- Read the result like a recruiter, not like a database
- Refine in plain language instead of starting over
- Common mistakes and how to fix them
- When to use filters or keywords instead
Most bad searches are not caused by the tool. They are caused by a brief that is either too thin ("developer Stockholm") or so long that every wish becomes a hard requirement and nobody is left. This guide shows how to find the middle ground, using how AI search in Essentiq actually reads your text.
How an AI candidate search reads your description
When you type a description into Essentiq AI Intelligence Search, the AI does not match your sentence word by word against profiles. It interprets the text and turns it into structured filters:
- Role or title, with synonyms added automatically (a search for Project manager also covers common variants)
- Location, a Swedish place plus a radius (40 km by default, up to 150 km, and up to 6 places in one search)
- Skills such as tools, methods or certifications
- Education: degree or field, school and graduation year
- Companies: current company and previous company
- Previous title, for experience the person had before their current role
You see these as chips under the text box before anything is searched. If part of your description could not be turned into a filter, Essentiq says so ("Couldn't be interpreted as search filters"), which is your cue to rephrase.
The search then covers over 4 million Swedish profiles and sorts the result: candidates who meet all your criteria come first, followed by people who are missing something, closest first. Every candidate card explains the match, for example Has worked as Project manager or Missing: skill.
The anatomy of a good search description
A good description has three layers. Write them in this order.
1. The role, in profile language
Use the title people actually put on their profiles, not the internal title from your job ad. "Customer Success Manager" will find more of the right people than "Client Happiness Lead". If the role has an established Swedish and English variant, either works: the AI adds synonyms, and you can add or exclude more afterwards.
Be specific about seniority only when it matters. "Senior" written into the role narrows the result to people who use that word in their title, which many strong candidates do not.
2. The place, and how strict it is
Name the town or city where the person should work or live: Göteborg, Linköping, Umeå. If the AI is unsure which place you mean, it asks you to choose one. After the first search you can decide how important the location is in the filters: Closest first ranks by distance, Only within the area removes everyone outside the radius.
3. Two or three must-haves
This is where most searches go wrong. Pick the requirements that genuinely rule someone in or out:
- a skill or tool (SAP, Siemens PCS 7, Excel, React)
- a type of education (civil engineering, nursing, economics)
- an industry signal, such as a previous employer
Everything else (personality, "driven", "team player", language nuance, salary expectations) cannot be read reliably from a profile. Leave it for the screening call.
Worked examples: weak vs strong descriptions
| Weak description | Why it struggles | Stronger description |
|---|---|---|
| Developer Stockholm | "Developer" is very broad, so the shortlist is huge and unranked by what you need. | Backend developer in Stockholm with experience of Java and Kubernetes |
| Driven salesperson with great energy who loves customers | No place, and the adjectives cannot be filtered. | Field sales representative in Göteborg with B2B experience in industrial products |
| Project manager, 10+ years, PMP, Prince2, SAFe, Jira, MSc, Swedish and English, construction, Malmö | Every wish becomes a criterion and almost nobody meets all of them. | Project manager in Malmö with a construction background. Add certifications in a second round. |
| Nurse | No location, so the search cannot run. | Registered nurse in Uppsala with intensive care experience |
| Controller who has worked at a big company | "Big company" is not a filter. | Financial controller in Stockholm who has previously worked at Ericsson or ABB |
Notice that the strong versions are not longer, they are sharper. A description of one or two sentences is usually ideal. The text box accepts much more, but more text means more criteria, and more criteria means fewer full matches.
Check the interpretation before you trust the result
Spend ten seconds on the chips before you search. Look for:
- The right title. If the AI picked "Sales manager" when you meant individual contributors, change it.
- The right place. Small towns can share names. Pick the correct one from the list.
- Skills that are really requirements. In the filters you can choose how skills are used: Rank (more matches rank higher), At least one required or All required. Start with Rank, tighten later.
- A graduation year that means what you think. The app reminds you that a graduation year may refer to a different degree than you intended, and many profiles do not list one at all.
Read the result like a recruiter, not like a database
The results open with the section Meets all criteria, followed by Missing something. Two habits make a big difference:
- Do not stop at the full matches. Profiles are often incomplete. Only about 30 % list their education, and many leave the skills field empty. Someone in "Missing something" may simply not have written the skill down. The app shows exactly what is missing, so you can judge quickly.
- Open the profile before you decide. The profile view shows career at a glance and an AI analysis of how the person fits your search, which is faster than reading the raw profile. See profile view and AI analysis.
Save the people you like to favorites in the project, and use Interesting and Not a fit while reviewing so the shortlist builds itself.
Refine in plain language instead of starting over
After the first search the text box changes to "Refine your search". Write what you learned, just as you would tell a colleague:
- "should also have Excel experience"
- "exclude interns and trainees"
- "only people who have worked as consultants before"
- "widen to Västerås and Eskilstuna"
When you mark a candidate as interesting or not a fit, you can also describe why ("What made the profile interesting?" / "What was missing?"), and the criterion is added to the search box for the next round. You can even start a new search from a strong profile with Start from this profile.
- The role is a title people actually use on their profiles
- A Swedish place is named, and you know whether location is a must or a preference
- No more than two or three must-haves in the first search
- Soft qualities and salary are left for the screening call
- You checked the chips (title, place, skills, graduation year) before searching
- You reviewed both Meets all criteria and the best of Missing something
- You refined with one change at a time, so you can see what each change did
Common mistakes and how to fix them
Too few results
You have probably turned preferences into requirements. Remove the least important skill, switch skills to Rank, widen the radius or set location to Closest first. If you excluded titles, check that you did not exclude a common variant by accident.
Too many results
Add one distinguishing requirement, such as a tool, a previous employer or a degree. Excluding titles helps too: − Intern or − Assistant removes a lot of noise for senior roles.
The right people rank low
Check the title chip. Only candidates with the role in their current title meet all criteria. If strong candidates have a different current title, add it as a synonym, or use previous title to find people who have done the job before.
When to use filters or keywords instead
AI search is the fastest way from a role description to a ranked shortlist. For niche technical terms, exact product names or a known person, the classic filter search is sometimes quicker: the Keywords field searches freely in the whole profile, and every word counts. Read Boolean search vs AI search for a deeper comparison, and the keywords filter article for the exact rules.
Search is free on every plan. When you have a shortlist, you can reach candidates with sequences on Pro, see pricing.
By Essentiq· Published
Keep reading
- ProductAI candidate search for SwedenDescribe the role in plain language. Essentiq searches over 4 million Swedish profiles, shows the best matches first and explains each one. Free to start.
- HelpWrite a search in plain languageDescribe who you are looking for in Essentiq AI Intelligence Search. The AI turns your text into search filters, searches and lets you refine in rounds.
- HelpWhat does “Meets all criteria” mean?Candidates who meet every requirement come first. The rest are grouped under Missing something, closest first, with chips showing what is missing.
- GuideBoolean search vs AI search for recruiters: when to use whichHow Boolean strings and AI candidate search differ, where each one wins, and a practical workflow that combines both for sourcing in Sweden.
- ArticleWhat does Essentiq cost? Prices and what is includedEssentiq has a free plan and Pro at SEK 999 a month for the first 3 months, then SEK 1,499. How credits, contact lookups and yearly billing work.
- ArticleWhat does LinkedIn Recruiter cost compared with Essentiq?LinkedIn Recruiter is sold per seat on a quote. How to compare the cost with Essentiq Pro, what both include and what is often left out of the sums.
