Job search
PDF or Website for a UX Job Portfolio? Choose the Format for the Review
An original fictional same-case PDF and website treatment plus six executed release-manifest checks distinguish format decisions, content consistency, actual export/access verification, and submission.
Choose the portfolio format the employer requests. When a UX or product design application permits both PDF and website, choose according to how the reviewer will encounter your work: a tailored, ordered document or a navigable collection with a direct case-study route. The case still needs clear evidence of your contribution; hiring teams do not share one universal format preference.
This comparison uses an entirely fictional case and two completed format decisions. LandOffer publishes this guide and offers job-search tools. Public guidance was checked October 21, 2026. No candidate portfolio was uploaded, submitted, published, or accessed; the examples are editorial specifications, not a rendered PDF or tested live website.

Use LandOffer’s recently posted jobs to identify the role first, then record the actual portfolio instructions before changing your materials.
Follow the requested route before comparing preferences
Read the application field and any linked instructions together. “Portfolio URL” may require a web address; “upload PDF” calls for a file. Do not assume that a PDF containing only a website link fulfills a document request, or that a hosted PDF satisfies every URL field. If the instructions conflict, ask for clarification through the designated channel before treating your preferred format as acceptable.
Real employer guidance also concerns the review itself. Atlassian’s design interview handbook encourages thoughtfully assembled work rather than jumping among preexisting PDFs or sharing entire design files. That is guidance about coherence and presentation; it is not a blanket rejection of PDF portfolios.
GitLab’s product designer preparation guidance says a formal presentation is not required and advises bringing materials together rather than sharing an entire design file. It asks about the problem, individual role, process, and results. Neither public page supplies a universal upload-size limit or proves what every application field accepts.
Record the requested format, size or page limits if specified, submission route, password instructions, and review context. An asynchronous review needs enough context to stand alone. A live discussion also needs material you can navigate while answering questions. Prepare the same evidence for the reading conditions of that particular review.
Compare the reading experience when both are permitted
NN/g’s portfolio guide describes web portfolios and PDF or slide-deck portfolios, including the ability to tailor document versions and the navigation of a web collection. Its format discussion asks whether the hiring manager specified a medium. The article was published in 2019; it remains accessible guidance, not evidence of a new 2026 hiring preference survey.
A PDF lets you set a default reading sequence and produce a role-specific selection. It can include a contents page and working links, but it still needs to explain the case without assuming the reviewer will follow external links. A saved document also freezes a version that you must deliberately update and name.
A website can offer a concise overview, project navigation, and a direct link to the relevant case. It also introduces dependencies: the address must remain usable, the intended reviewer must be able to access the page, and essential content must not depend on an unavailable embed or an unexplained password. Those are checks to perform on your actual site, not failures inherent to every website.
You can tailor either medium to the role. A PDF can omit unrelated projects; a website can lead with the relevant case or use a carefully maintained case-specific route. Keep a revision record so a tailored selection does not change the underlying dates, contribution, or result.
Worked asset one: the same case, two complete editorial treatments
Fictional designer Leah Cho is applying for a product design role involving multilingual content discovery. Her invented MuseumRoute case concerns selecting an audio-guide language. She designed the selector prototype and explanatory copy; a content editor owned translation accuracy, and an engineer advised on audio availability. There was no launch or measured usage improvement.
The evidence set has six required elements: context, contribution, problem evidence, alternatives, decision, and limitations. Leah’s fictional review notes say visitors could not reliably infer a language from a flag symbol. They do not establish a representative usability finding. Her design decision was to use written language names and show whether audio was available, accepting a longer list rather than implying every language had an audio track.
Here is the completed six-page PDF content specification. The page count is Leah’s editorial choice, not an employer requirement.
Page 1 — Project and role. “MuseumRoute: choosing an audio-guide language. Prototype exploration, April–May 2026. I designed the selector flow and availability copy. The content editor owned translation accuracy; engineering advised on available audio assets. This case shows a design decision and its limits, not a shipped product.”
Page 2 — The problem. “A flag-only selector asked visitors to infer language from a symbol. In our fictional review notes, that interpretation was uncertain. I focused the design question on choosing understandable labels while making audio availability explicit. The notes were exploratory, not a representative test of all museum visitors.”
Page 3 — Options considered. “Option A retained a compact flag grid. Option B used written language names with availability text. Option A occupied less space, but the symbol-to-language assumption remained. Option B required a longer list and more scrolling. I kept both options visible so the trade-off could be reviewed.”
Page 4 — Decision and illustration caption. “I selected the text-label list for the next prototype. Each choice states the language and whether an audio track is available. The accompanying schematic would show a supported option beside an unavailable one. It is a proposed treatment; it does not prove translation quality or complete language coverage.”
Page 5 — Collaboration and unfinished work. “I prepared the revised selector and copy. The editor still needed to review the labels, and engineering still needed to validate the availability data contract. The next evaluation would ask whether people can select their intended language and explain whether audio is available. That evaluation has not occurred.”
Page 6 — Result and limits. “The completed output is the revised prototype direction with a documented rationale. No implementation, accessibility validation, representative usability result, or production improvement is established. I would revisit the design if content coverage, list length, or evaluation evidence changes the trade-off.”
The website treatment carries the same facts with a different entry route:
Home-page project card: “MuseumRoute — making audio-guide language choices understandable. My role: interaction design and copy exploration. Prototype only. Read the case to compare a compact flag grid with a text-label list.”
Case opening: “Visitors need to choose a language and understand whether an audio track exists. I designed a revised selector for this fictional prototype; translation approval and implementation remained with other owners. Start with the problem, or jump to the decision and limitations.”
Navigation and case body: The case has six labeled anchors: Context, My contribution, Problem evidence, Alternatives, Decision, and Limits. The filled content above is adapted into those sections: the two options remain adjacent, the availability caption stays beside its schematic, and the unfinished editorial and engineering checks remain visible before the closing result. “Decision” does not link directly to an isolated final screen without its caveat.
Closing summary: “This work produced a proposed selector and rationale, not a verified language-access outcome. Translation review, availability validation, and evaluation of the revised flow remain open.”
These are completed writing treatments, not proof that a file rendered or a website loaded. The invariant is the case’s factual scope. Navigation and page order may change; Leah’s contribution, project dates, rejected option, and untested status do not.
Keep essential evidence inside the chosen route
A PDF should not require a separate prototype link to explain the central decision. Include enough stills, captions, and context to make the comparison understandable if that link fails. A website likewise needs a readable account if a video player or design-file embed is unavailable. An optional interactive artifact can add detail after the core case is understood.
Check the specific limit in the application rather than adopting a generic megabyte target. If you reduce document size, inspect the exported result for readable labels and diagrams. A file’s existence is not evidence that its type, size, links, or reading order meet the request. This article produced no PDF, so it makes no file-size or rendering claim for Leah’s specification.
For a website, test the actual intended address without relying on your signed-in session. Confirm that the case opens, navigation reaches the right sections, and the reader can find your contribution and limitations. Record what you observed and when. A public home page does not prove a linked case, attachment, or embedded artifact is accessible.

Worked asset two: completed format and release decisions
Leah receives two fictional instructions, not real employer messages. Review A requests one PDF and specifies a file limit in the application. Review B permits either a PDF or a website and asks for the multilingual interaction case. No actual submission occurred.
For Review A, Leah chooses the six-page document treatment. She removes the unrelated homepage project selection from the application package. The format decision is complete, but release is held until the actual PDF is exported and checked against the stated limit, including readability and links. A website substitution would still leave Review A’s PDF instruction unmet.
For Review B, Leah initially considers the direct website case. Her release record marks its signed-out access as unverified, so she does not describe it as tested or publicly readable. She selects the tailored PDF as the primary preparation route because its proposed content gives the reviewer one ordered story. That is a contextual choice, not a conclusion that all PDFs are easier to access. Actual export checks still remain necessary.
An original local Python exercise checked six declared release manifests. These were invented metadata records, not real files or network observations:
- A1: PDF-only instructions paired with a website candidate: held for format mismatch.
- A2: PDF candidate containing only a link to the case: held because essential evidence was external.
- A3: PDF specification with all six evidence elements and the current case revision: content checks passed; physical export checks remained pending.
- B1: Website candidate with unverified signed-out access: held for the access unknown.
- B2: Website candidate whose result text claimed a launch absent from the case record: held for inconsistent evidence.
- B3: Tailored PDF specification with the current six elements and matching limitations: content checks passed; physical export checks remained pending.
The executed fixture returned two content-consistent specifications and four held candidates, with the reasons saved for each record. It also rejected a mutation that removed the limitations element and one that changed the expected revision. It did not open a PDF, visit a portfolio, test a password, measure file bytes, or certify accessibility.
The completed release note reads: “Review A: PDF selected. Review B: tailored PDF selected while website access remains unverified. A3 and B3 are the content baselines. Both remain preparation artifacts, not released or submitted files. Next release gate: verify the actual export and each instruction-specific constraint.”
Maintain one factual case record across versions
Give each rendition a revision identifier and record what changed: project selection, order, caption length, or visual crop. Keep substantive changes synchronized. If a prototype has since been evaluated, update the evidence and limitations in every rendition before claiming a result; do not update only the more visible website.
Permission is a separate decision. A password-protected page, a private file link, or blurred material does not itself establish that you may share underlying employer or participant information. Use material you are authorized to show and exclude restricted content. The MuseumRoute example is wholly fictional and needs no real confidential source material.
Before release, inspect the actual reviewer route from entry to closing limitation. Use the format requested, and make the core case understandable without an unexplained extra step. Neither this decision nor a clean local manifest check establishes that an application has been submitted.
Find a relevant role through LandOffer’s current listings, record its portfolio instructions, select the complete evidence route, and verify the actual artifact before release.


