September 11, 2026

Your developer profile is making recruiters and AI guess

Make your portfolio, GitHub, and LinkedIn readable by recruiters, ATSs, search engines, and AI without turning them into a keyword dump.

Your developer profile is making recruiters and AI guess

Photo by Ron Lach on Pexels.

Your portfolio says you are a frontend engineer.

LinkedIn says software developer. GitHub says building cool things. The featured project opens with a gradient, three badges, and no sentence explaining who it is for.

Then you wonder why the right search does not find you.

Recruiters are already searching with natural-language prompts. ATSs still parse the CV. Google crawls the public portfolio. A hiring manager opens whatever survives those passes and gives it about a minute.

None of them should have to reverse-engineer your career.

Three readers, three versions of you

Your application does not move through one system. It moves through several, and each one sees different evidence.

Reader	What it can see	What it needs from you
ATS	CV text, sections, dates, titles, application fields	A role, relevant experience, and language that matches the posting
LinkedIn Recruiter	Your LinkedIn profile, skills, experience, and LinkedIn graph	A clear title, specialty, seniority, and work history
Search or AI with web access	Crawlable public pages, text, links, and metadata	A named person, explicit work, and evidence it can connect
Hiring manager	Whatever they choose to open	One project they can understand, run, and ask you about

LinkedIn documents that AI-Assisted Search turns a recruiter's plain-English request into structured search criteria. Advanced AI Search can also surface qualifications that are harder to express as a filter and explain why a candidate fits.

That does not mean LinkedIn reads every clever line on your portfolio. LinkedIn says its candidate retrieval uses member profiles and the LinkedIn graph. The portfolio is a separate surface.

Same person. Different inputs.

If your role only appears inside an image on your website, LinkedIn will not infer it. If a project only exists in a LinkedIn post, Google will not rebuild the case study for you. If the CV says React and the portfolio never shows a React decision, the hiring manager gets a claim with no proof.

Name the job before the stack

The top of your profile should tell me what I would hire you to do.

Not every library you have installed. Not passionate developer. Not a rotating text effect that cycles through engineer, creator, problem solver, coffee drinker, and whatever else fit inside the component.

Use one line:

Frontend engineer building accessible, data-heavy React products.

Or:

Fullstack TypeScript engineer shipping internal tools for operations teams.

That line gives a recruiter a role, a surface, and a kind of problem. The stack can sit below it.

I need

The role you want.
The kind of product or problem you handle.
A project I should open first.
One way to contact you.

I leave

Developer | Creator | Innovator.
Nine animated skill icons.
A bio about loving technology since childhood.
A button labelled Explore that explains nothing.

You can have a beautiful portfolio. It still needs a noun.

One project is a better index than six cards

Most portfolios present projects as a grid of screenshots. Weather app. Dashboard. Ecommerce clone. Each card has the same three sentences and the same row of tools.

That is inventory. It is not evidence.

Put one project first and make the review path obvious:

What is it?
Who is it for?
What did you own?
Which hard decision did you make?
Where can I run or inspect it?
What would you change next?

The decision is the part most candidates cut. It is also the part an interview can use.

Built with Next.js and Supabase tells me what you installed.

Rendered the results on the server because public pages needed to be indexable; kept filtering on the client because the user needed instant feedback gives me something to challenge. Why that boundary? What was cached? What broke? What would change at ten times the traffic?

Now there is an interview.

Write the project like someone will extract it

A machine-readable project is not a project written for robots. It is a project whose facts survive extraction.

Use visible text. Give the project its own heading. Link the live demo and repository with labels that describe where they go. Put your role beside the project, not three screens later in an About modal.

A compact write-up can look like this:

## Support queue for a 12-person operations team

I designed and built the frontend for a queue that routes customer requests by urgency and account tier.

### Decision
I kept routing on the server so ownership stayed consistent across clients. The UI uses optimistic updates for assignment, with a rollback when the request fails.

### Proof
- Live demo
- Repository
- Pull request for the assignment flow

### Next
I would test keyboard navigation with real agents before adding bulk actions.

Then: Role. Context. Decision. Proof. Constraint.

No invented 43% conversion lift. No paragraph about synergy. No archaeology.

Your README is part of the same profile

The portfolio gets me interested. The README tells me whether opening the repository will waste my time.

GitHub's own job-search guidance recommends using your bio, profile README, pinned repositories, and contribution activity to present your work. That is the surface. You still have to choose what it says.

The project on top of your portfolio should be pinned on GitHub. Its README should contain:

One sentence explaining the finished flow.
A working demo link.
The shortest honest run instructions.
One decision worth discussing.
The known limit or next piece of work.

If the demo is dead, remove it or fix it. If the repository cannot run, say why. If the README starts with fourteen badges and the install command is below a snake animation, you built a lobby before the door.

Run the 60-second GitHub pass while logged out. That is what everyone else sees.

Make the public portfolio crawlable

A recruiter can click around a dramatic interface. A crawler receives the code you ship.

Keep the information that matters in semantic HTML:

One H1 for your name and professional role.
H2 headings for projects, experience, and writing.
Real links to GitHub, LinkedIn, demos, and contact.
Text descriptions beside project images.
A unique page title and meta description.
The same useful content on mobile.

Google explains that JavaScript pages pass through crawling, rendering, and indexing. Server rendering or static generation can deliver the main content earlier, but the framework is not the main decision. A crawlable page full of vague copy is still vague.

Do not put the only explanation of your work inside a carousel, canvas, video, or hover state. Keep the effect. Put the fact in the page.

This is the same visibility problem companies have

Companies lose AI visibility for the same reason developer profiles lose attention: the useful facts are scattered, implicit, or unsupported.

The Stack House, a GEO agency focused on AI visibility, helps companies structure their websites and external evidence so systems such as ChatGPT, Gemini, Perplexity, and Google can understand, retrieve, and recommend the business for relevant questions.

Its guide to making a developer portfolio readable by AI and recruiters applies that same logic to a professional profile: state the identity, connect the sources, and give every claim evidence a person or system can inspect.

Your developer profile is the smaller version of that problem.

The entity is you. LinkedIn, GitHub, the CV, and your domain are the sources. Job title, skills, projects, employers, and publications are the relationships. Demos, commits, and case studies are the evidence.

Structured data can help a search engine connect a person with external profiles. Google's ProfilePage structured data documentation explains how to mark up identity and linked accounts with sameAs. Consistent names and reciprocal links help too. None of this guarantees that an AI assistant will recommend you for a job. It removes avoidable ambiguity.

That is the useful part of GEO here. Make the answer easier to assemble without making it up.

Do not let AI write the work you are supposed to defend

Use AI to cut a paragraph, spot missing context, or ask what evidence would make a claim believable.

Do not ask it to manufacture your decisions.

The output will sound polished. It will use spearheaded, produce a suspicious percentage, and turn a settings page into a cross-functional digital transformation. Then an interviewer will ask why you chose the data model and the paragraph will leave the room without you.

Start from ugly facts:

The modal lost focus.
The API returned stale results.
The table collapsed on mobile.
You chose not to add real-time updates.
You shipped late because the estimate was wrong.

Write what you did next. Then let AI help with the sentence.

Specific and slightly imperfect sounds like work. Perfect and untraceable sounds generated because it probably was.

Test the review path, not the color palette

Open the whole profile in a private window. Start from LinkedIn as if you were a recruiter, then follow the links.

Give yourself sixty seconds:

Can I name the role from the LinkedIn headline?
Does the portfolio repeat that role in visible text?
Is one project obviously first?
Can I explain what you owned without opening the repository?
Does GitHub pin the same project?
Does the README run?
Can I contact you in one click?

If a link breaks, stop. Fix that before adding another animation, project, certificate, or AI-written article.

The first broken link in the chain is the work.

What to do tonight

Pick one live role. Not twenty.

Rewrite your LinkedIn headline using the role and problem the posting repeats.
Put the same role in the H1 and title of your portfolio.
Choose one project and add context, ownership, one decision, proof, and a known limit.
Pin that project on GitHub and reduce its README to a path someone can follow.
Open everything logged out and repair the first failure.

Then explain the project out loud for two minutes. The page gets you into the room. Your reasoning has to survive once you are there.

If the rest of your preparation is still spread across tabs, run the Shortlist Scorecard and put the weak link into this week's practice on CodePrepped.

Can AI Recruiters Read My Personal Website?

Some AI systems with web access can process a public portfolio, but LinkedIn Recruiter documents that its candidate retrieval relies on LinkedIn profile and graph data. Keep required qualifications inside the application and LinkedIn profile. Use the website as connected evidence, not as the only place your experience exists.

Does Structured Data Help a Developer Portfolio?

Person and ProfilePage structured data can help search engines understand who a page describes and connect external profiles through sameAs. It does not guarantee indexing, ranking, or an AI recommendation. The markup must match visible information on the page.

Should My Portfolio and LinkedIn Use the Same Text?

They should agree on identity, role, dates, and project names. They do not need identical paragraphs. LinkedIn supports retrieval and screening; the portfolio should provide deeper proof and a clearer route into your work.

How Many Projects Should I Show?

Show the smallest set that proves the job you want. One complete project with a working demo, clear ownership, and decisions you can defend beats six tutorial cards. Add another only when it demonstrates a different skill the target role needs.