The builder is the easy part. Almost nobody fails at the software, and almost everybody fails at deciding what goes on the page.
A no-code AI site builder will produce a complete portfolio site from a prompt: Durable claims a complete site in 30 seconds from three questions[15], Wix generates pages, copy and visuals from a plain-language description[7], and Webflow calls its output “a functional site and a foundational design system that you can shape”[10]. None of them can pick your three best projects or write your results. Build the page on three established frameworks instead: Jobs-to-be-Done for what the visitor is really there for[4], Nielsen Norman Group’s F-pattern eyetracking research for where their eyes actually go[1], and STAR (Situation, Task, Action, Result) for case studies that land[6].
I’ve hired freelancers, agencies and in-house marketers for about a decade, and I can tell you exactly what happens at the moment of decision. Two candidates look equally credible on paper. One has a link. I click it, and within roughly ninety seconds I either think “yes, this person has done my exact problem before” or I go back to the other tab.
That ninety seconds is the entire job of a portfolio site. It isn’t a lead generation channel, and anyone selling you a personal website as a traffic strategy is selling you something. What it does is win the last mile. Your LinkedIn profile is formatted the way LinkedIn wants it and gets reordered whenever the feed changes; your resume is a list of claims with no evidence attached. A portfolio page is the one surface where you control the order things are read in.
There’s a second reason that’s harder to ignore. The World Economic Forum’s Future of Jobs Report 2025 found employers expect 39% of workers’ core skills to change by 2030, “significant ongoing skill disruption” even though it’s down from 44% in 2023[17], drawing on over 1,000 companies across 22 industries and 55 economies[18]. If nearly two-fifths of what you’re good at will shift inside five years, a static resume describing a job title is a poor container for it.
One honest note first. I went looking for a credible statistic on what share of hiring managers actually visit a candidate’s personal website, because it would have made a great opening number. Every version I chased led back to a marketing blog quoting another marketing blog, so I’m not going to give you a percentage I can’t stand behind. What I’ll give you instead is what every hiring manager I know does: we click the link when there is one, and we form an opinion fast.
Here is what separates a portfolio that gets replies from one that just exists: it was built around a decision someone else is making, not around a story you want to tell. Three established frameworks handle that, and none was invented as a personal branding framework or for portfolio sites at all. That’s exactly why they work.
Jobs-to-be-Done comes out of two Harvard Business Review pieces: Tony Ulwick’s outcome-driven version in January 2002[5], and the better-known 2016 article by Clayton Christensen, Taddy Hall, Karen Dillon and David Duncan[4]. In plain English: people don’t buy products, they hire them to get a job done. As HBR’s own summary puts it, “when we buy a product, we essentially ‘hire’ it to help us do a job,” and jobs “are never simply about function; they have powerful social and emotional dimensions”[4].
Apply that to your portfolio page. The person who lands on it has been handed a problem by their own boss, and they’re doing one job: decide, quickly, whether this person can fix my particular problem without creating a new one for me.
That emotional half is the whole game and almost every portfolio ignores it. Your visitor doesn’t just want a competent freelancer; they don’t want to look foolish in three weeks for having recommended you. So the homepage has two jobs at once: prove capability, and reduce the risk of choosing you. Concretely, that means a headline naming the problem you solve rather than your job title. Compare “Hina Mian, Marketing Consultant” with “I fix B2B email programmes that have stopped converting.” The first makes the visitor do the translation work.
This one is measured rather than theorised, which is why I trust it more than most design advice. Nielsen Norman Group recorded how 232 users looked at thousands of web pages and found a dominant reading pattern in three parts: a horizontal sweep across the upper content area, a second shorter sweep further down, and a vertical scan down the left side[1]. Together they roughly trace an F. NN/g revisited it eleven years later and found it alive on desktop and mobile, and mirrored in right-to-left languages[2].
The implications are blunt and directly usable. Users won’t read word by word. “The first two paragraphs must state the most important information.” And “start subheads, paragraphs, and bullet points with information-carrying words,” because people “read the third word on a line much less often than the first two words”[1].
Here’s the nuance design blogs lose. NN/g is explicit that the F-pattern is bad and that it’s a symptom of your formatting rather than a layout to design toward: it “is the default pattern when there are no strong cues to attract the eyes towards meaningful information”[2]. NN/g’s later work names a better target, the layer-cake pattern, where eyes move across headings then drop into the text under the one that matters, “by far the most effective way in which users can scan pages” short of reading every word[3]. On a portfolio page that means a hero line saying what you do, plus project headings that name a result rather than a client. “Cut CAC by 34% for a Series A fintech” earns a fixation. “Project Two” does not.
On the Z-pattern, since it always comes up: it’s a design convention borrowed from print and a fine rule of thumb for a sparse hero. But NN/g’s eyetracking work doesn’t report a Z among its measured text-scanning patterns, which are F, spotted, layer-cake, commitment, marking and bypassing[2,3]. Use it to arrange a hero block if it helps. Don’t cite it as research, because I can’t find the research.
STAR is a behavioural interview technique that adapts almost perfectly to a case study. The UK Government’s National Careers Service defines the parts as: “Situation, the situation you had to deal with; Task, the task you were given to do; Action, the action you took; Result, what happened as a result of your action and what you learned from the experience,” and says explicitly that it works in a CV or application form, not only an interview[6].
Most portfolio case studies are all Action. Pages of process, mood boards, “we ran a discovery workshop,” then a screenshot, then nothing. No Situation, so the reader can’t tell whether it was hard. No Result, so they can’t tell whether it worked.
Before: “Rebranded and relaunched the email programme for a B2B SaaS client. Built new templates, rewrote the welcome sequence, and set up segmentation in HubSpot.”
After (STAR): “Situation: a 40,000-contact B2B list with a 12% open rate and no segmentation, one newsletter going to everyone. Task: get email contributing pipeline again within one quarter. Action: cut the list to 22,000 engaged contacts, split it into four behavioural segments, rebuilt the welcome sequence as five emails triggered by what people had clicked. Result: open rate 12% to 31% over eleven weeks, and email went from 4% to 19% of sourced pipeline. What I’d do differently: I cut the list before the re-engagement flow was ready, losing about 3,000 contacts I could probably have saved.”
That last sentence is the one people leave out and the one that does the most work. Someone who names their own wrong turn reads as a person who has actually run things, and it answers the “will this make me look bad” half of the job directly.
I checked every one of these against the vendor’s own current pages while writing this, because the feature sets move fast and half of what’s published about them is stale. Wix’s original ADI tool, for one, was retired on 10 November 2024 and replaced by something that works completely differently, which a lot of “best AI website builder” articles still haven’t noticed[7]. There are three distinct categories of tool behind any no-code portfolio website, and it’s worth knowing which one you’re buying.
You describe your business in plain language and the tool generates a whole site. Wix says “by simply describing your vision, the AI generates a complete site with industry-specific logic, tailored copy and high-quality visuals already wired in,” after which you move between prompting its Aria agent and drag-and-drop editing[7]. Squarespace’s Blueprint AI uses guided questions instead of free text, generating “a customized website draft tailored to your answers, including page structure, content, and images”[8], inside a wider Design Intelligence suite that adds AI copy and brand identity tools[9]. Durable makes the boldest claim of the lot: a “complete, professional website in just 30 seconds” from three questions[15].
Notice how carefully the two developer-adjacent ones word it. Webflow promises to “turn a prompt into a fully customized, production-ready website,” then describes the output as “a functional site and a foundational design system that you can shape to meet your needs”[10]. Framer’s agent “creates editable pages, sections, copy, and visuals directly in your project,” and Framer states plainly that “AI helps you move faster while you make the final design decisions”[11]. Framer also meters AI by credits, 500 of them free, with a paid plan needed for a custom domain[12].
Canva Websites is not a prompt-to-site product, whatever the roundups tell you. Canva’s own page describes choosing from “designer-made one-page and multipage website templates” and scopes the AI to “stunning visuals, instant copy, and interactive widgets”[13]. Canva Code gets closer to prompt-to-page, but it’s a premium feature drawing on a monthly AI allowance[14]. Carrd is the honest outlier: no AI, doesn’t pretend to have any, still one of the fastest ways to get a single-page portfolio online, free for up to three sites with custom domains on Pro from $19 a year[16]. For plenty of people reading this, that’s the correct answer and you can stop shopping.
| Tool | What it generates | Free tier reality | Still done by hand |
|---|---|---|---|
| Wix AI Website Builder | Complete site from a prompt, copy and visuals “already wired in”[7] | Free to start; Premium plan for a custom domain or payments[7] | Every design decision, via Aria or drag-and-drop |
| Squarespace Blueprint AI | Draft from guided questions: page structure, content, images[8] | Free to use, paid account to publish; desktop only[8] | Text, images, layouts, site styles[8] |
| Webflow AI site builder | “A functional site and a foundational design system”[10] | Paid Site plan to connect a domain[10] | Add, remove, reorder sections; site-wide styles[10] |
| Framer | Editable pages, sections, copy and visuals from a prompt[11] | 500 one-off AI credits; paid plan for a domain[12] | Final design calls, including breakpoints[11] |
| Durable | Complete site in a stated 30 seconds from three questions[15] | Free to create, subscription to publish[15] | Layout, images, headings, fonts[15] |
| Canva Websites | Templates; AI does visuals, copy, widgets, not the whole site[13] | Free on my.canva.site, up to five live sites[13] | Assemble the pages yourself |
| Carrd | Nothing automatically. Templates only, no AI[16] | Free for three sites; Pro from $19/year adds domains[16] | All of it, about an hour for one page |
Capability and pricing claims taken from each vendor’s own current product page, checked August 2026. This is a summary of vendor statements, not independent benchmark testing.
This is where the quality difference actually gets made, and it’s the category people skip. General assistants like ChatGPT, Claude or Gemini are what you use before a single pixel exists: pulling the Situation and Result out of a rambling project description, generating the sceptical questions a hiring manager would ask about your numbers, cutting a 400-word write-up to 90 without losing the evidence. Worth keeping a prompt library for, rather than rewriting the same instructions every time.
My honest view after testing a lot of these: every builder produces a competent-looking site and none produce a memorable one, because what makes a portfolio memorable is a judgement call about what to leave out. No model knows which three of your eleven projects are worth showing.
At Future Factors we teach adoption as a multiplication problem, and it’s the single most useful thing we’ve built:
Tool x Workflows x Behavior = AI-powered professional
Writing it as multiplication rather than addition matters. If any one of the three is zero, the result is zero, and you can’t compensate for a missing workflow by buying a better tool. Portfolio sites are the cleanest demonstration of this I’ve seen, because the failure is so visible.
Tool is the builder. The easy variable, the one everyone optimises, worth maybe 20% of the outcome. Wix, Framer, Carrd, doesn’t matter much.
Workflow is the recurring update habit: where finished work goes, when you write it up, what triggers a change to the site. Most people have none of this. No folder, no trigger, no calendar entry.
Behavior is treating the site as a living asset rather than a project with an end date. The difference between “I built my portfolio” and “I keep my portfolio current.”
Set Workflow and Behavior to zero and you get one of two outcomes. I’ve watched both happen dozens of times.
Failure one: it never ships. The person stays permanently at the Tool stage, comparing builders for three weeks, signing up for two free trials, concluding they should probably learn Webflow properly first. No live site eighteen months later. Procrastination wearing the costume of research.
Failure two: it ships once and rots. More common, and more painful because the work got done. Somebody spends a focused weekend, publishes something genuinely good, sends the link around, then never touches it again. Two years on it shows a job title they left, a project from a previous industry, and a contact form pointing at an email they’ve stopped checking. That site is doing active harm, because a stale portfolio reads as a signal about how you’d handle a client’s work.
So, honestly: what will make you update this site three months from now? If the answer is “I’ll remember,” you’ve set Workflow to zero and the maths is done. Set a trigger before you open any builder. It takes four minutes and it’s the highest-return four minutes in the whole process.
This sequence assumes an afternoon, and it deliberately puts the builder at step four. Opening a builder before you know what goes on the page is how people end up with a beautiful site that says nothing.
The sequence described in this section. This is Future Factors’ recommended workflow, not measured data.
Write this down literally: “Someone lands on my site because they need to decide whether I can ______ for them, and they’re worried that ______.” If you can’t fill the second blank, you don’t yet know your visitor and everything downstream comes out generic.
Then three projects. Not eleven. The curation is the signal, and three sharp case studies beat nine vague ones because nine tells the reader you couldn’t decide what your best work was. Write all four STAR parts longhand, plus the “what I’d do differently” line[6]. If an NDA means no numbers, use a directional result (“cut approval time from three weeks to four days”) rather than dropping the R.
That last instruction isn’t optional. Left alone, a model will invent a plausible percentage to fill a gap, and nothing torches your credibility faster than a made-up number you can’t defend in a call.
Expect a starting point, because that’s what the vendors themselves promise. Squarespace says Blueprint AI “gives you a custom starting point tailored for your brand, but you can edit your AI-powered website anytime”[8]. Webflow’s flow is generate, then “add some structure,” then “make it your own”[10]. Framer will ask whether you want tablet and phone breakpoints added, which tells you responsive behaviour isn’t automatically finished[11].
This is the twenty minutes that does more than any theme choice, working from NN/g’s own antidotes to F-pattern scanning[2]:
Then open the site on your phone and read only the left-hand third of the screen. If that alone doesn’t say what you do and what you achieved, you haven’t finished. On whether a page converts the person reading it, our guide to AI-assisted conversion rate optimization goes deeper.
Check what your plan allows: Squarespace requires a paid account to publish at all[8], Durable requires a subscription to publish[15], and Wix, Webflow, Framer and Carrd all need a paid tier for a custom domain[7,10,12,16]. Shipping on a free subdomain beats a perfect site that stays in draft. Then, in the same sitting, do the four-minute thing from the last section. Don’t tell yourself you’ll do it tomorrow.
Roughly in order of how often I see them. The first accounts for more failures than the rest combined.
The homepage describes you instead of the job. “Strategic marketing leader with a passion for data-driven storytelling” gives the reader nothing to act on, and it’s near enough word for word what an AI builder generates from a job title and nothing else. Name the problem you solve.
Everything you’ve ever made is on there. That reads as an archive rather than a case. Three strong pieces with a through-line beat eleven that show you’ll take any work going.
Process without outcome, or results with no baseline. Six screens of wireframes, a screenshot, then the next project: that’s the Result in STAR going missing[6]. And “grew traffic by 200%” means nothing until you know it went from 400 to 1,200 monthly visits.
A wall of text with no formatting. NN/g found people scan in an F precisely when a page has “a ‘wall of text’ but no bolding, bullets, or subheadings”[2]. You’re not fighting attention spans, you’re fighting an absence of cues you failed to provide.
The AI copy went up unedited. Everyone can spot it, because it’s the same three sentence shapes on every AI-generated site online. Rewrite the hero and the about section in your own words at minimum. The same problem shows up on resumes, and our guide to writing a resume with AI without it sounding generic covers the fix.
No obvious next step, or an NDA wall with nothing behind it. I’ve clicked through portfolios where I wanted to hire the person and had to hunt for an email address. And confidentiality is real, but “I can’t share this” with no substitute is a dead page. Anonymise it: “a 300-person logistics company” plus situation, action and a directional result gets most of the value with none of the risk.
The rebuild-every-two-years cycle exists because people let the site drift so far from reality that editing feels harder than starting over. Which by then it genuinely is. The fix isn’t discipline, it’s making each update small enough that it never becomes a project.
This is the habit that changes everything, and I learned it the annoying way. When a campaign wraps, spend ten minutes right then screenshotting the dashboard and writing down the before and after numbers, while you still have access and still remember the baseline.
I once lost a genuinely excellent result because I went back for the numbers eight months later and my access to the client’s analytics had been revoked. I had the story and no evidence, so it never went on the site.
A calendar reminder on its own gets dismissed. Pair it with event triggers so updating attaches to something that already happened: a project ends with a result worth showing, you change roles or target client, or someone asks a question your site should have answered. Then add the backstop, which is the four-minute setup I keep mentioning. Open your calendar now, create a recurring 20-minute event called “portfolio refresh,” and paste the site link into it. That’s the Workflow variable handled.
Cap the case studies at three or four permanently, too. When a new one earns its place, the weakest comes down. That keeps every update near twenty minutes and stops the slide into an eleven-project archive nobody reads.
Given that employers expect 39% of workers’ core skills to shift by 2030[17], editing four times a year rather than rebuilding every two isn’t fussiness. It’s the only version that keeps up.
If I had to reduce this whole guide to one action: before you compare a single builder, write the sentence describing the decision your visitor is making, and write one project in Situation, Task, Action, Result with a real number in it. Do those two things and any of these tools will produce something good. Skip them and the best builder on the market will hand you a handsome page that gets closed in nine seconds.
Our guide to building a LinkedIn personal brand with AI covers the profile most people see first, and if the portfolio is aimed at a job search rather than clients, using AI to prepare for the interview is the natural next step once the link starts getting clicked.
Yes, and it’s genuinely a matter of hours rather than weeks now. Wix generates a complete site from a plain-language description of what you want, with copy and visuals included[7], Squarespace’s Blueprint AI produces a draft including page structure, content and images from a few guided questions[8], and Durable claims a complete site in 30 seconds from three questions[15]. Every one of those vendors describes the output as a starting point you then edit: Webflow calls it “a functional site and a foundational design system that you can shape,” and Framer states that “AI helps you move faster while you make the final design decisions”[10,11]. The coding was never the hard part. Deciding what belongs on the page is.
Jobs-to-be-Done says people don’t buy products, they hire them to get a specific job done. Harvard Business Review’s summary of the 2016 Christensen, Hall, Dillon and Duncan article puts it as: “when we buy a product, we essentially ‘hire’ it to help us do a job,” and notes that jobs “are never simply about function; they have powerful social and emotional dimensions”[4]. Tony Ulwick published the outcome-driven version of the same idea in HBR back in January 2002[5]. On a portfolio site, the visitor’s job is to decide quickly whether you can solve their particular problem without creating a new one for them. The emotional half of that job, not wanting to look foolish for recommending the wrong person, is why your page should reduce perceived risk as well as prove capability.
Three separate categories, and it helps to know which you’re buying. Prompt-to-site builders (Wix, Squarespace Blueprint AI, Webflow’s AI site builder, Framer, Durable) generate a whole site from a description or a few questions[7,8,10,11,15]. Template builders with AI features bolted on (Canva Websites) give you designer-made templates plus AI-generated visuals, copy and widgets rather than a full generated site[13]. And Carrd has no AI at all, which for a one-page portfolio is often fine: free for up to three sites, with custom domains on Pro from $19 a year[16]. The third category is the general assistants (ChatGPT, Claude, Gemini) that you use before you open a builder, to pull the results out of your own project write-ups.
The UK Government’s National Careers Service defines STAR as: “Situation, the situation you had to deal with; Task, the task you were given to do; Action, the action you took; Result, what happened as a result of your action and what you learned from the experience,” and states it works in a CV or application form, not just an interview[6]. Most portfolio case studies are pure Action: process, wireframes, workshops, screenshots. Adding the Situation tells the reader whether the problem was hard, and adding the Result tells them whether your work moved anything. Always include the baseline (“12% open rate to 31%” rather than “tripled engagement”), and add a short line on what you’d do differently. That last line reads as experience and it’s the one nearly everyone omits.
Twenty minutes quarterly as a backstop, plus event triggers: whenever a project ends with a result worth showing, whenever your focus or target client changes, and whenever someone asks a question your site should have answered. The reason to keep the cadence short is that the alternative is a full rebuild every two years, which happens because the site drifts so far from reality that editing feels harder than starting over. It’s also worth noting that the World Economic Forum’s Future of Jobs Report 2025 found employers expect 39% of workers’ core skills to change by 2030[17]. The habit that makes all of this easy: capture your screenshots and before-and-after numbers at the end of each project, while you still have dashboard access.
This guide is built on ten years of hiring marketers and reviewing their portfolios, combined with three established frameworks traced to primary sources: Jobs-to-be-Done via Ulwick’s 2002 and Christensen et al.’s 2016 Harvard Business Review articles, F-pattern reading behaviour via Nielsen Norman Group’s original 2006 eyetracking study and its 2017 and 2019 follow-ups, and the STAR method via the UK Government’s National Careers Service. Every tool capability and pricing claim was checked against the vendor’s own current product page in August 2026 rather than taken from training knowledge or third-party roundups. A statistic on the share of hiring managers who visit candidates’ personal websites was deliberately excluded because it could not be traced past marketing blogs citing one another. The Z-pattern is discussed as a design convention rather than research, because no measured eyetracking evidence for it was found.
Weekly tips. Real examples. Practical help for busy professionals.
We care about your data, check out our privacy policy.