Loading
Technical SEO Audit & Optimization Plan
Manan Devs SEO Audit
A full audit of manandevs.vercel.app against the goal of winning freelance and Shopify client work — on-page, technical, structured data, keywords, off-page entity signals, content gaps, and booking conversion. Findings are ordered by what actually moves the needle, not by category.
- Audited
- 5 Sep 2026
- Stack
- Vite 8 · React 19 · react-router 7 · Tailwind 4
- Routes
- 4 + 404
- Findings
- 18
Headline finding — verified live
The site serves an empty HTML shell. Every route ships the homepage's title and canonical tag.
I fetched the live site. manandevs.vercel.app/ returns a body containing only <div id="root"></div> — no headings, no copy, no project names. And because vercel.json rewrites every path to the same index.html, /works returns the homepage's <title> and <link rel="canonical" href="https://manandevs.vercel.app/">. I confirmed this on the live /works URL.
react-helmet-async corrects the tags after React hydrates, but a canonical tag is a directive read from the delivered HTML. Right now the raw HTML tells search engines that /works, /capabilities and /contact are all duplicates of the homepage — while sitemap.xml simultaneously submits them as four distinct pages. Those two signals directly contradict each other.
Everything else in this audit is downstream of that. Fixing headings, schema, and keywords on a page that ships no HTML is optimizing text most crawlers never receive.
Scorecard
The seven areas you asked about, rated on what is live today.
| Area | Status | Assessment |
|---|---|---|
| Rendering & indexability | Solved | Fixed (QW-01). All 11 routes now prerender to static HTML at build time via scripts/prerender.mjs, each delivering its own fully rendered body and its own Helmet-resolved title, description and canonical. The catch-all rewrite in vercel.json is gone, so real files are served and an unknown URL returns a genuine 404 from 404.html. |
| On-page & headings | Solved | Fixed (QW-02). /contact leads with a real <h1> — "Hire a Full-Stack Next.js Developer" — and all ten <p>-as-heading call sites are now proper <h2>/<h3> elements. Every route has exactly one <h1> and a real outline. Typography now logs a dev error when a heading variant is used without an explicit as, and the test suite fails on it, so the bug class cannot come back. |
| Titles & descriptions | Solved | Fixed. Prerendering put every title and description in the delivered HTML, so they no longer depend on a crawler executing JS. Lengths were then brought inside the truncation limits — all 11 titles now sit at 44–58 characters and all descriptions at 129–153, verified against the built output. |
| Core Web Vitals readiness | Strong | Genuinely good engineering: build-time image manifest with intrinsic dimensions, blur placeholders, lazy loading, font-display: swap, immutable asset caching, split vendor chunk. One flaw — the LCP route is code-split behind lazy(). |
| robots / sitemap / canonical | Solved | Fixed. sitemap.xml is now generated by scripts/prerender.mjs from the same route list that emits the HTML, so it cannot drift — 11 URLs with a real lastmod, and changefreq/priority dropped because Google ignores them. Canonicals are per-route and verified by curl. robots is emitted once per page by <Seo>, replacing the site-wide tag that collided with the 404 page’s noindex. |
| Structured data | Solved | Fixed. FAQPage is now generated from the array the FAQ component renders and ships only on the two routes that show it; ProfilePage was added to the homepage alone; every case study carries CreativeWork/WebApplication plus BreadcrumbList; sameAs gained Behance and Cal.com. All 12 pages parse as valid JSON-LD. One residual: Figma is still absent because its URL cannot be verified — tracked in the audit log. |
| Booking conversion | Solved | Fixed (QW-03). Every CTA — navbar, hero, the CTA block and each case-study footer — is a real <a href="https://cal.com/abdulmanan"> that upgrades to a Cal.com popup once the embed script loads, so the booking survives a blocked script. /contact now opens with the inline calendar, with the form beneath it as "prefer to write first?" and WhatsApp and email as tertiary options rather than the only channel. |
| Off-page & entity signals | Partial | Partly fixed. The on-site half is done: rel="me" is on every outbound profile link and Behance is now linked from the footer and declared in sameAs. Still Partial because the remainder is off-site work no code change can do — confirming the Figma handle, making handles consistent across platforms, and adding the reciprocal link back to this site from each profile. |
| Content depth | Solved | Fixed (ST-02). The six projects are now six /works/<slug> case studies — roughly 900 words each, structured problem → what I built → the technical decisions and why → outcome — with their own canonical, OG image, CreativeWork/WebApplication and BreadcrumbList schema. That takes the site from four indexable pages to eleven. Still outstanding and tracked separately: no blog, no dedicated About page, and no Shopify work (ST-03). |
Quick wins
Ship this week. Each is hours, not days, and each is a real defect rather than a tuning opportunity. Effort tags are honest estimates for someone who knows this codebase.
QW-01
Prerender the four routes to static HTML~4 hours
vite.config.ts · package.json
This is the fix for the headline finding, and it's smaller than it sounds. You have four static routes and no per-request data — that's the exact case static prerendering was built for. You do not need to migrate to Next.js.
vite-react-ssg is the drop-in for a react-router app: it crawls your route table at build time and emits a real works/index.html, contact/index.html and so on, each with its own Helmet-resolved title, description and canonical baked into the delivered HTML. The client bundle then hydrates on top. Nothing about your component code changes.
Do this
Install vite-react-ssg, convert src/app.jsx's <Routes> into an exported route array, swap the createRoot call in src/main.jsx for the SSG entry, and change the build script. Then remove vercel.json's catch-all rewrite so real files are served, keeping a rewrite only as the 404 fallback. Verify with curl -s https://manandevs.vercel.app/works | grep -i canonical — it must return the /works canonical, not /.
QW-02
Give /contact an H1, and turn ten fake headings into real ones~1 hour
src/components/common/typography.jsx:63 and 10 call sites
Your Typography component defaults to as = 'p'. So <Typography variant="h1"> renders a paragraph that looks like a heading. Ten call sites do exactly that — Why Us, Featured Projects, Testimonials, About, FAQ, CTA, Tech Stack, Performance, Features, and the contact page's own title.
The consequences: /contact has no H1 and no headings whatsoever; the homepage outline jumps H1 → H3 → H3 with no H2 in between; and every section title — the copy carrying your keywords — is semantically invisible.
Do this
Add as="h2" to each section heading, and as="h1" to contact-form.jsx:90. Then guard the class of bug: make variant without an explicit as throw in dev, or default as to the element matching the variant. Rewrite the contact H1 to carry intent — "Hire a Full-Stack Next.js Developer" beats "Let's build something amazing together."
QW-03
Put the Cal.com link on the site — it isn't there~2 hours
navbar.jsx:53 · hero.jsx:51 · cta.jsx:45 · contact.jsx
I grepped the entire codebase for cal.com. Zero matches. Every button reading "Free Strategy Consultation" or "Book a Free Strategy Call" navigates to /contact, whose form serialises into a WhatsApp deep link to a +92 number.
That intake channel loses exactly the clients you're targeting. A US or EU founder evaluating three developers will not open WhatsApp to message an unknown international number — they'll book whoever offers a calendar. You already own that calendar and it isn't linked.
Do this
Point the navbar CTA straight at https://cal.com/abdulmanan. Add the Cal.com inline embed as the first element on /contact, above the form — booking for people who are ready, the form for people who are not. Keep the WhatsApp path as a third option rather than the only one. Add the Cal URL to sameAs.
QW-04
Add Behance, Figma and Cal.com to sameAs~15 min
index.html — Person node
sameAs is how you tell Google that the GitHub account, the LinkedIn profile and this site are one entity. Yours currently lists three URLs and omits half your professional presence — including Behance, which is the only place your design work lives.
Do this
Replace the sameAs array with all six URLs (see the structured data section below), and add rel="me" to the matching outbound links in footer.jsx and about.jsx.
QW-05
Fix the soft 404~30 min
src/pages/not-found.jsx
not-found.jsx has no <Helmet>, so a nonexistent URL inherits the homepage title and the homepage canonical — telling Google that /anything-at-all is a valid duplicate of your homepage. It also returns HTTP 200. And the page spends its only <h1> on the digits "404".
Do this
Add a Helmet block with <meta name="robots" content="noindex, follow"> and no canonical. Make the visible "404" a decorative <div aria-hidden="true"> and promote "Page Not Found" to the H1. Once prerendering lands (QW-01), serve a genuine 404 status from Vercel. Add links to /works and /contact so the page recovers the visit.
QW-06
Reconcile the FAQ schema with the visible FAQ~20 min
index.html — FAQPage node vs. src/components/sections/faq.jsx
The JSON-LD declares four questions ("What services does Abdul Manan offer…", "Why choose Next.js and React…"). The rendered FAQ component contains five entirely different ones ("What services do you offer?", "How long does a project usually take?"…). Structured data must describe content actually on the page — a mismatch is a manual-action risk, not a style issue.
Worse, the block sits in index.html, so it is served on /works and /capabilities where no FAQ is rendered at all.
Do this
Generate the FAQPage node from the same faqData array the component renders, and emit it only on routes that render the FAQ. Calibrate expectations: Google restricted FAQ rich results to government and health sites in August 2023, so this earns no SERP feature — you are fixing it to remove risk, not to gain stars.
QW-07
Generate sitemap.xml at build time~45 min
public/sitemap.xml · scripts/
Every lastmod reads 2026-03-20 — six months old as of today. A hand-maintained sitemap always drifts, and stale lastmod values train crawlers to ignore the field. changefreq and priority are ignored by Google entirely; they are harmless noise.
Do this
Add a build step that emits the sitemap from the route list with real dates — ideally each route's last git commit date, which is both accurate and free. Drop changefreq and priority.
QW-08
Rewrite four alt attributes that carry no information~20 min
capabilities-grid.jsx:82 · tech-stack.jsx:86 · hero.jsx:59 · testimonials.jsx:127
Alt text is mostly good here — about.jsx:31 and the project screenshots are solid. Four are weak:
capabilities-grid.jsx:82andtech-stack.jsx:86usealt={item.title}, duplicating the heading directly beside the image. A screen reader hears the same phrase twice and the alt describes nothing.hero.jsx:59—alt="Abdul Manan"on an avatar decorating a button. The button already has a label, so this should bealt="".testimonials.jsx:127—alt={testimonial.name}on an Unsplash stock photo. See ST-04; the deeper problem is the testimonials themselves.
Do this
Add a distinct alt field to the techStack and capabilities data arrays describing what each image shows. Set the hero avatar to alt="".
QW-09
Label the contact form fields~20 min
src/components/sections/contact-form.jsx:200–260
All five inputs are placeholder-only. Placeholders vanish on focus, aren't announced reliably, and fail WCAG 3.3.2. This is your primary conversion surface — it's the one form on the site that turns traffic into money.
Do this
Add a visually-hidden <label htmlFor> per input, or a visible floating label. Add autoComplete (given-name, family-name, email) so mobile browsers autofill it — that measurably lifts completion. Change projectType from a free-text input to a <select> with your actual service lines, which both qualifies the lead and tells you which service attracts demand.
Structural fixes
This month. Each is a day or more, and each raises the site's ceiling rather than patching a defect.
ST-01
Buy a real domain before you build a single backlink1 day + DNS
manandevs.vercel.app is a subdomain of a host on the Public Suffix List. It inherits nothing from vercel.app, and it reads to a prospective client as a deploy preview rather than a business.
The sequencing matters more than the fix. Every backlink you place over the next quarter is an asset — and if you migrate the domain afterwards, you spend that equity on 301s and lose a slice of it permanently. Do this before the off-page work, not after.
Do this
Register manandevs.com or manandevs.dev. Point it at the same Vercel project, set the apex as primary, 301 the .vercel.app host to it, and update every absolute URL: the canonical and OG tags in index.html, all four Helmet canonicals, every @id in the JSON-LD graph, sitemap.xml, robots.txt, and manifest.json. Then update the link on all six external profiles the same day.
ST-02
Split the six projects into six case-study URLs2–3 days
src/pages/works.jsx · src/data/projects.js
Six projects currently share one URL, so the site offers Google exactly four indexable pages. Each project deserves /works/<slug> with its own title, description, canonical, OG image and CreativeWork schema — that is six new entry points targeting six different long-tail queries instead of one page trying to rank for all of them.
Your project data already carries title, description, image and stack — the routing and metadata come almost free. What takes the time is writing real case studies, and that is the part that converts.
Do this
Add a slug to each entry in projects.js, add a /works/:slug route, and register the slugs with the prerenderer so all six emit static HTML. Structure each page: the problem → what you built → the technical decisions and why → the outcome. The technical-decisions section is what makes it an experience-and-expertise signal rather than a screenshot gallery. Aim for 800–1,200 words. Then link each Behance project and each GitHub repo to its case study.
ST-03
You sell Shopify and have no Shopify workvaries
The meta description promises "Shopify builds," and the JSON-LD offer catalog lists "Shopify Development & Customization." Not one of the six portfolio projects is Shopify — they are Next.js dashboards, a React marketing site, an AI media tool, a video SaaS, and an agentic-infrastructure site.
That gap is a conversion problem before it's a ranking problem: a Shopify lead lands, finds no Shopify evidence, and leaves. It's also the reason a Shopify keyword strategy can't work yet — there is nothing on the site for those queries to match.
Do this
Pick one and commit. Either build one real Shopify proof asset — a custom theme section, a headless Shopify + Next.js storefront demo, a migration teardown — and give it a case study; or drop Shopify from the copy and schema until you have work to show. The headless route suits you: it is genuinely a Next.js job and it targets "headless Shopify developer," a far less crowded term than "Shopify developer."
ST-04
The testimonials are fabricated, and that's an E-E-A-T liability1 week, mostly waiting
src/components/sections/testimonials.jsx:4–70
Six testimonials attributed to Alex Morgan, Sofia Ramirez, Daniel Kim, Emily Carter, Michael Chen and Olivia Wilson, each with a stock Unsplash portrait. Several are written in the plural — "the team delivered" — on a solo developer's site.
I'd be doing you a disservice not to flag this directly. A prospect who reverse-image-searches one avatar has a reason to distrust everything else on the page, including the "35+ creators and founders" claim in the hero. And if you ever add Review or AggregateRating markup on top of invented reviews, that is a direct violation of Google's structured data policies — the kind that draws a manual action.
Do this
Email your last five real clients today asking for two sentences and permission to use their name and company. Meanwhile, replace the section with something true: outcome cards ("Landing page rebuild — Lighthouse 62 → 98, LCP 4.1s → 1.2s") carry more weight with a technical buyer than a generic quote anyway. Add Review schema only once real, attributable reviews exist. LinkedIn recommendations are the fastest source — they are public, verifiable, and you can link to them.
ST-05
Don't code-split the landing route~30 min
src/app.jsx:10 · navbar.jsx:35 · footer.jsx:66
Two things are working against your LCP, and both are cheap to fix.
First, Home is behind lazy() like every other route. That puts a serial waterfall in front of your largest contentful paint: HTML → entry chunk → vendor chunk → then the home chunk → then the H1 paints. The entry route should be statically imported; only secondary routes benefit from splitting.
Second, the nav and footer use raw <a href> for internal routes. Those are crawlable, which is good — but each click triggers a full document reload, throwing away the SPA and re-running every Core Web Vital.
Do this
Import Home directly; keep lazy() for the other three. Swap internal <a href> for react-router <Link>, which still renders a real crawlable <a href> — you keep the SEO benefit and regain client-side navigation. Leave the /#anchor links as plain anchors.
Structured data
Your existing @graph is better than most portfolios ship — @id references are used correctly and the Person ↔ ProfessionalService link is sound. Three things to change.
1 — Complete the sameAs array
This is the single highest-value schema edit and it takes five minutes. Every URL here should also link back to the site, so the relationship is confirmable in both directions.
"sameAs": [
"https://github.com/manandevs",
"https://www.linkedin.com/in/manandevs/",
"https://www.behance.net/abdulmanan263",
"https://www.figma.com/@manan44",
"https://cal.com/abdulmanan",
"https://www.fiverr.com/manann_pro"
]2 — Add a ProfilePage wrapper
ProfilePage is the type Google uses to understand that a page is about a person rather than merely mentioning one. Emit this on the homepage — or better, on a dedicated /about route (see content gaps).
{
"@type": "ProfilePage",
"@id": "https://manandevs.com/#profilepage",
"url": "https://manandevs.com/",
"mainEntity": { "@id": "https://manandevs.com/#person" },
"dateCreated": "2026-03-20T00:00:00+05:00",
"dateModified": "2026-09-05T00:00:00+05:00"
}3 — One CreativeWork per project page
Emit this on each /works/<slug> page from ST-02. WebApplication is the more precise type for the SaaS builds; use CreativeWork for the marketing sites. The author reference is what attaches each project to your Person entity — that accumulation of attributed work is the actual E-E-A-T signal.
{
"@context": "https://schema.org",
"@type": "WebApplication",
"@id": "https://manandevs.com/works/clipza#project",
"name": "Clipza — SaaS Video Sharing & Recording Platform",
"url": "https://manandevs.com/works/clipza",
"sameAs": "https://clipza-peach.vercel.app/",
"applicationCategory": "MultimediaApplication",
"description": "Instant screen recording and video management platform with
cloud storage uploads, real-time security and global delivery.",
"image": "https://manandevs.com/images/projects/clipza.webp",
"author": { "@id": "https://manandevs.com/#person" },
"creator": { "@id": "https://manandevs.com/#person" },
"isPartOf": { "@id": "https://manandevs.com/#website" },
"keywords": "Next.js 16, React 19, MongoDB, Better Auth, Arcjet",
"datePublished": "2026-08-01"
}Also worth adding
BreadcrumbList on every project page once ST-02 ships — it is the one item here that reliably produces a visible SERP change, replacing the raw URL with Home › Works › Clipza.
Validate before you ship. Google's Rich Results Test and Schema.org's validator disagree on edge cases; run both. And once you are on a real domain, register Search Console — none of this is measurable until you do, and you currently have no visibility into what is indexed at all.
Keyword strategy
Grouped by whether you can realistically win them, not by search volume. A new site on a Vercel subdomain with no backlinks cannot compete for the head terms — so the plan is to own the winnable tail now and build toward the rest.
| Keyword | Tier | Target page | Note |
|---|---|---|---|
| manan devs · abdul manan developer | Own now | Homepage | Brand. Should already rank; verify once Search Console is live. |
| next.js saas developer for hire | Winnable | Homepage H1 + /capabilities | Your strongest fit — four of six projects are exactly this. |
| freelance full stack developer pakistan | Winnable | /about | Geo-modified terms are far less contested. Needs a real About page. |
| react dashboard developer freelance | Winnable | /works/fern-dashboard | Case study is the ranking asset here, not the services page. |
| headless shopify next.js developer | Winnable | New Shopify page | Much thinner competition than "shopify developer". Blocked on ST-03. |
| hire next.js developer | 6–12 mo | Homepage | Toptal, Arc, Upwork and Clutch hold page one. Needs domain + links first. |
| next.js developer for hire | 6–12 mo | Homepage | Same competitive set. Include in copy; don't build the plan around it. |
| shopify developer portfolio | Reframe | — | Low volume and mostly navigational — people searching it want a template, not a hire. Target "hire shopify theme developer" instead. |
| next.js vs react for saas | Blog | /blog | Informational, links well, and demonstrates the expertise you are selling. |
| how much does a saas mvp cost | Blog | /blog | High commercial intent disguised as research. Strong lead magnet. |
Where each keyword goes
One primary term per page, in the H1, the title tag, the first 100 words, and one H2. Right now every page's copy circles the same "full-stack Next.js" cluster, so your four pages compete with each other. Once ST-02 splits the projects out, give each case study a distinct stack-plus-outcome phrase and the internal cannibalization resolves itself.
The honest framing: for a freelancer at your stage, organic search is not the main lead channel and won't be for a year. Your leads come from Fiverr, LinkedIn, and referrals. SEO's job right now is to convert those visitors — someone who has been given your name and is checking you out — and to slowly accumulate the long-tail entries. Optimize for that first, and treat head-term rankings as a later dividend.
Off-page & entity signals
You have five external profiles with real domain authority behind them. Three currently link to the site and none of them link consistently. Do ST-01 (the domain) first.
Make the identity graph bidirectional
For a person rather than a local business, the NAP equivalent is a consistent name + brand + URL triple, confirmed in both directions. Site links out with rel="me"; each profile links back to the same canonical URL. That mutual confirmation is what lets Google merge six accounts into one entity. Your handles are currently inconsistent, which weakens the association:
| Platform | Handle | Links back? | Action |
|---|---|---|---|
| GitHub | manandevs | Yes | Add a profile README linking the site. Set the homepage field on every pinned repo to its case study URL. Pin the portfolio repo itself. |
| manandevs | Yes | Highest-authority link you control. Website slot + Featured section + one case study republished as an article per month. | |
| Behance | abdulmanan263 | Not linked | Rename to manandevs if available. Every project description should link its case study URL, not the homepage. |
| Figma | manan44 | Not linked | Your most underused asset. Publish one Community file — a Tailwind design-token kit or a SaaS dashboard UI kit — and it earns a dofollow profile link plus recurring discovery traffic. |
| Fiverr | manann_pro | Partial | Link the portfolio from the gig descriptions. Keep it in sameAs — it's a real credential. |
| Cal.com | abdulmanan | Not linked | Add your site URL and headshot to the booking page so it reads as one brand, not a stray calendar. |
Link building that is actually worth your time
- Cross-post with
canonical_url. Publish on your own blog first, then republish to Dev.to and Hashnode withcanonical_urlpointing home. You get their reach and their audience; your domain keeps the ranking credit. - Open-source your components. Your
Imagecomponent — build-time manifest, intrinsic dimensions, inlined blur placeholder — is a genuinely good piece of work. Ship it as a standalone package with a README linking your site. Useful packages earn links passively for years. - Answer in public. Stack Overflow and GitHub Discussions in the Next.js and Tailwind ecosystems. Nofollow, but real referral traffic from exactly the technical buyers you want.
- Get listed. Awwwards, Land-book and Godly for the design side; the Vercel and Tailwind showcase submissions for the technical side.
What to skip: paid directories, guest-post networks, and anything offering “50 DA40+ backlinks.” They are a liability, not a shortcut.
Content gaps
Ranked by return on the effort each takes.
| Missing | Effort | Why it earns its place |
|---|---|---|
| Six case-study pages | 2–3 days | Highest return of anything on this list. Turns four indexable pages into ten, gives each project a long-tail query to own, gives Behance and GitHub somewhere specific to link, and makes dwell time real — a prospect reading a technical breakdown stays minutes, not seconds. Covered in ST-02. |
A dedicated /about page | half a day | About is currently a homepage section with no URL of its own, so there is no page to attach ProfilePage to, nothing for "abdul manan developer" to rank, and nothing to link from six external profiles. Include your actual history, location, and how you work. |
| Real testimonials | 1 week | Removes the credibility risk in ST-04 and unlocks legitimate Review markup. Mostly waiting on replies, so start the emails today. |
| A blog | ongoing | The only sustainable way to build topical authority and earn links without asking. One substantial post a month beats four thin ones. Write what you already know: the image-optimization pipeline in this repo, why you chose Vite over Next.js here, Better Auth vs. NextAuth from the Clipza build. |
| Per-service pages | 1–2 days | /capabilities lists five services on one URL, so none of them can rank. Split at least the two you most want work in — SaaS development and Shopify — into their own pages with their own keyword, testimonial, and case study. |
| Pricing or engagement page | half a day | Not for rankings — for qualification. "Projects start at $X" filters out the leads that waste your calls, and ranks for "how much does a next.js developer cost." |
Conversion: the Cal.com path
Traffic you cannot convert is a vanity metric. Your CTA placement is already good — the issue is entirely where the buttons point.
What is there now:
- Navbar — "Free Strategy Consultation", persistent and above the fold. Good placement, hidden below
lg, routes to/contact. - Hero — "Free Strategy Consultation" with your avatar and "Book Your Free Strategy Call" beneath it. Strong, personal, and it promises a call it does not deliver.
- CTA block — appears on all four pages. Good repetition.
- Contact form — five fields, then a WhatsApp deep link to a
+92number.
So four separate calls to action say “book a call,” and none of them books a call. The arrangement to replace it:
Navbar
Every page, always visible
- Change label to "Book a Free Call"
- Link straight to
cal.com/abdulmanan - Show it below
lgtoo — mobile is likely most of your traffic
Hero
First screen
- Cal.com popup embed, not a route change
- Keep the avatar — a face on a booking CTA lifts conversion
- Replace the subtext with proof: "Usually replies within a day"
Contact page
Highest-intent visitors
- Cal.com inline embed at the top
- Form below it, framed "Prefer to write first?"
- WhatsApp and email as tertiary options
Case studies
After the proof lands
- Contextual CTA at the end of each
- "Want something like this? Book a call."
- Highest-converting position on the site
Configure the booking itself
Name the event type for the buyer, not the calendar — "Free 30-min Project Consultation," not "30 Min Meeting." Add two or three intake questions (project type, timeline, budget range) so calls arrive pre-qualified. Set generous availability across US and EU business hours; you are in PKT and your target clients are not. And add a ?utm_source=portfolio parameter so you can tell booked calls apart from cold ones.
Keep WhatsApp — it is genuinely effective for regional clients and for people who do not want a meeting. Just stop making it the only door.
Sequence
The order matters more than the list. Domain before backlinks; rendering before content; Search Console before anything you intend to measure.
Week 1
Quick wins — foundations
- Register the real domain and 301 (ST-01)
- Verify in Search Console + Bing Webmaster Tools
- Fix the headings and the contact H1 (QW-02)
- Link Cal.com everywhere it is promised (QW-03)
- Complete
sameAs+rel="me"(QW-04) - Fix the 404 (QW-05), alt text (QW-08), form labels (QW-09)
Week 2
Make the site indexable
- Prerender all four routes (QW-01)
- Verify per-route canonicals with
curl - Generate the sitemap at build (QW-07)
- Reconcile the FAQ schema (QW-06)
- Un-split the home route, switch to
<Link>(ST-05) - Request indexing for all four URLs
Weeks 3–6
Build the surface area
- Six case-study pages (ST-02)
CreativeWork+BreadcrumbListper page- Dedicated
/about+ProfilePage - Real testimonials replace the stock ones (ST-04)
- Update all six profiles to deep-link case studies
Quarter
Compound the gains
- Resolve the Shopify gap either way (ST-03)
- Split SaaS and Shopify service pages
- Publish the Figma Community file
- One substantial blog post per month
- Package and publish the
Imagecomponent - Review Search Console monthly and re-prioritize
One thing worth saying plainly
Your site sells technical SEO and Core Web Vitals optimization as a service, and lists server-side rendering as the reason to choose Next.js. The site making that argument is a client-rendered SPA that delivers an empty <div> to every crawler, with one canonical URL shared across four pages.
Any technical prospect who checks — and the ones worth having will check — sees that. Which makes QW-01 worth doing for the sales argument alone, before you count a single ranking. The rest of the engineering here is careful work; this one gap undercuts all of it.
Audit log
A running record of what has been checked, what was fixed and what is still open. Entries are never rewritten — when something is resolved it moves from Outstanding to Fixed with a note on how. Last run: 6 Sep 2026.
Static analysis of src/ plus verification against the built dist/ output served through a static server that replicates the host's file-serving semantics, because vite preview rewrites every path to / and hides exactly the problems this audit is looking for.
Fixed (19)
CriticalEvery per-page tag — title, description, canonical, robots, OG, Twitter — was rendered inside `<body>`, not `<head>`
6 Sep 2026
Found while confirming the Search Console verification tag reached every route.
react-helmet-asyncdoes not populate its SSR context under React 19 —helmetContext.helmetcomes backundefined— soheadFor()injected an empty string at the<!--app-head-->marker. React 19 handles document metadata natively instead, and because this app renders a fragment into#rootrather than the whole document, React had no<head>to hoist into and emitted all of it inline in the body. Browsers hide the damage (React hoists on hydration), but the delivered HTML is what crawlers read, and<link rel="canonical">and<meta name="robots">outside<head>are ignored outright. This silently undid the per-route canonical work.Resolved: The prerenderer now extracts
<title>,<meta>,<link>and JSON-LD from the rendered markup and injects them at the head marker, keeping the Helmet-context path for if it ever starts working again. Verified across all 12 pages: each has exactly one title, description, canonical and robots tag inside<head>, and zero metadata left in<body>.Files:
scripts/prerender.mjsMinorEarlier audit passes verified tag presence per document rather than per `<head>`
6 Sep 2026
The checks used
curl | grepacross whole documents, which counts a canonical in<body>as a pass. That is why three consecutive passes reported per-route canonicals as verified while every one of them was in the wrong half of the document. A check that cannot fail the thing it is testing is not a check.Resolved: All metadata assertions now slice the document at
</head>and count within it, and additionally assert zero leakage into<body>. The scorecard and log entries above were re-verified under the corrected method.Files:
audit method — no source changeModerateSitemap `lastmod` was the build date, so every URL claimed to change on every deploy
6 Sep 2026
Found on re-audit, and it was a defect introduced by the earlier fix for the stale hand-maintained sitemap. Stamping
new Date()onto all 11 URLs means a deploy that changed one case study told crawlers that all 11 pages changed. That is the same failure as a six-month-stalelastmodwearing the opposite mask: the field carries no information either way, and crawlers learn to ignore it.Resolved: Each route now declares the source files that decide its content (
getStaticRoutesinentry-server.jsx), and the prerenderer takeslastmodfromgit log -1over those paths, falling back to the build date where git history is unavailable. Verified it returns 2026-09-05 and 2026-07-25 for untouched files rather than today.Files:
scripts/prerender.mjs,src/entry-server.jsxModerate`/works` carried no collection or breadcrumb schema despite being the hub for all six case studies
6 Sep 2026
Every case study had
BreadcrumbList, but the page linking to all of them described itself only through the site-widePerson/WebSitenodes. Nothing told a crawler that this URL is a collection, which six items it contains, or where it sits in the hierarchy — on the one page whose entire job is to funnel authority into the six pages the site most wants ranked.Resolved: Added a
CollectionPagewith anItemListnaming all six case studies in order, plus theBreadcrumbListplacing/worksunder the homepage. Generated from theprojectsarray, so it cannot drift from the rendered cards.Files:
src/pages/works.jsxModerateThe Arbiris screenshot is 1200 × 7637 and renders roughly 4,450px tall
6 Sep 2026
It is a full-page capture with a 1:6.4 aspect ratio. On
/worksamax-h-100keeps it contained, but the case-study<figure>sets no height limit, so the intrinsic dimensions make the browser reserve a box around 4,450px tall at typical widths. That is a large layout-shift and scroll-depth problem on one of the six pages the site is trying to rank.Resolved: Added
max-h-150 object-cover object-topto the case-study screenshot incase-study.jsx— one class change, capping the rendered box at 600px while keeping the intrinsic width/height attributes that prevent layout shift.Files:
src/pages/case-study.jsxMinorTestimonial avatars use a real person’s name as alt text on stock photography
6 Sep 2026
alt={testimonial.name}labels an Unsplash stock photo as a named individual. The alt problem is cosmetic; the credibility problem underneath it is not, and it is the reason this is worth resolving rather than patching.Resolved: Set
alt=""intestimonials.jsxso the avatar is exposed as decorative — the name is already adjacent in text. The underlying credibility issue (stock photos standing in for real clients) is unchanged and still tracked under Not Yet Reviewed.Files:
src/components/sections/testimonials.jsxMinor`og:image:width` and `og:image:height` are no longer emitted
6 Sep 2026
They were dropped from
index.htmlwhen the per-page OG tags moved into<Seo>, because case studies use their own screenshots and the homepage dimensions would have been wrong for them. Some scrapers use the pair to reserve space before fetching the image.Resolved:
<Seo>now reads intrinsic dimensions from the build-time image manifest and emits the pair per page, falling back to 1200×630 for the default OG image.Files:
src/components/common/seo.jsxMinorNavigation uses raw `<a href>` for internal routes instead of `<Link>`
6 Sep 2026
Every nav click triggers a full document load rather than a client-side transition. Crawlability is unaffected — these are real hrefs, which is what matters for SEO — but it discards the router and makes navigation measurably slower, which reaches SEO indirectly through engagement.
Resolved: Route links in
navbar.jsxandfooter.jsxnow render<Link to>; hash links stay plain anchors because they target sections, not routes. Verified the built HTML still emits realhrefattributes, so nothing changed for crawlers.Files:
src/components/layout/navbar.jsx,src/components/layout/footer.jsxModerateOpen Graph card for the Arbiris case study was a 1200 × 7637 sliver
6 Sep 2026
Found while emitting
og:imagedimensions: each case study passes its own screenshot as the social image, and social platforms crop to roughly 1.91:1. A full-page capture at 1:6.4 would have been cropped to an unrecognisable strip on every share.Resolved: Case studies now check the screenshot’s aspect ratio against the manifest and fall back to the site OG image when it is below 0.5:1. Only Arbiris trips it; the other five keep their own screenshot.
Files:
src/pages/case-study.jsxModerateNo `ProfilePage` node, and no `rel="me"` on the outbound profiles
6 Sep 2026
Two halves of the same entity claim were missing.
ProfilePageis the type that tells Google a page is *about* a person rather than merely mentioning one, andrel="me"is the on-site half of the reciprocal link thatsameAsasserts from the schema side.Resolved: Added a
ProfilePagenode tohome.jsx— emitted through<Seo>so it claims the homepage only, not all 11 pages the way the oldFAQPagenode did — andrel="me"plus the missing Behance link to the footer profiles.Files:
src/pages/home.jsx,src/components/layout/footer.jsxCriticalFAQ structured data described questions that appear nowhere on the site
6 Sep 2026
The
FAQPagenode inindex.htmldeclared four questions ("What services does Abdul Manan offer at Manan Devs?", "Why choose Next.js and React…") while the rendered FAQ contains five entirely different ones. Because the node lived in the shared template it was also served on all 11 pages, including/worksand every case study, where no FAQ is rendered at all. Structured data must describe content visible on the page; a mismatch this size is a manual-action risk rather than a style problem.Resolved: Removed the hand-written node from
index.htmland generate the schema insidefaq.jsxfrom the samefaqDataarray the component renders, emitted via Helmet. It now ships only on the two routes that actually render the FAQ (/and/contact) and cannot drift from the visible questions.Files:
index.html,src/components/sections/faq.jsxModerateTwo conflicting `robots` meta tags on the 404 page
6 Sep 2026
index.htmlcarried a site-widecontent="index, follow, max-snippet:-1…"andnot-found.jsxaddedcontent="noindex, follow", so404.htmlshipped both. Crawlers resolve conflicting directives to the most restrictive, so the behaviour happened to be correct — but the page was sending two contradictory instructions, and the next page needing a noindex would inherit the same ambiguity.Resolved: Removed the global tag from
index.html.<Seo>now emits exactly onerobotsdirective per page — the full index/follow string by default,noindex, followwhen the route passesnoindex.Files:
index.html,src/components/common/seo.jsxModerateEvery case-study title tag overran the SERP truncation limit
6 Sep 2026
All six ran 77–98 characters ("AI Media Editing Platform Case Study — WebAssembly, Next.js & Client-Side Processing | Manan Devs" is 97). Google cuts around 60, so the brand and half the differentiating terms were being dropped from the visible result on the six pages the site most needs to rank.
Resolved: Rewrote all six to 44–54 characters, leading with the primary term and keeping the brand suffix.
Files:
src/data/case-studies.jsModerateEight meta descriptions exceeded the 160-character display limit
6 Sep 2026
The six case studies (190–208 chars),
/works(162) and/seo-audit(229) were all being cut mid-sentence in results, wasting the snippet that decides whether the result gets clicked.Resolved: Rewrote all eight to sit between 140 and 155 characters, each ending on a complete clause.
Files:
src/data/case-studies.js,src/pages/works.jsx,src/pages/seo-audit.jsxModerate`sameAs` omitted half the professional profiles
6 Sep 2026
sameAsis how a search engine confirms that the GitHub account, the booking page and this site are one entity. It listed three URLs while the site links a Cal.com calendar site-wide and the design work lives on Behance — neither was declared, so neither contributed to entity consolidation.Resolved: Added
https://www.behance.net/abdulmanan263andhttps://cal.com/abdulmananafter confirming both return 200. Figma is deliberately still absent — see the outstanding item.Files:
index.htmlModerateFooter FAQ link was relative and broke on every non-home route
6 Sep 2026
The link was
href="#faq"with no leading slash, so from/worksor any case study it resolved to/works#faq— a page with no FAQ section. An internal link that lands on nothing wastes crawl equity and dead-ends the visitor.Resolved: Changed to
/#faqso it always resolves to the homepage FAQ.Files:
src/components/layout/footer.jsxModerateNo `trailingSlash` policy, so each URL was reachable two ways
6 Sep 2026
With the prerendered directory layout,
/worksand/works/both resolved and returned identical HTML. The canonical tag pointed at one form, so this was mitigated rather than harmful — but two live URLs per page splits any links that arrive at the other form.Resolved: Set
"trailingSlash": falseinvercel.jsonso the host 308-redirects the slashed form to the canonical one.Files:
vercel.jsonMinorNo `og:image:alt` on any page
6 Sep 2026
Social cards were shipping an image with no text alternative, which affects accessibility on every platform that surfaces it and is a documented Open Graph field.
Resolved: Added
og:image:altandtwitter:image:altto<Seo>with a sensible site default; case studies pass an alt describing their own screenshot.Files:
src/components/common/seo.jsx,src/pages/case-study.jsxMinorEight images used their adjacent heading as alt text
6 Sep 2026
capabilities-grid.jsxandtech-stack.jsxboth passedalt={item.title}, duplicating the heading sitting directly beside the image. A screen reader announces the same phrase twice and the alt describes nothing about the image itself.Resolved: Added a distinct
imageAltfield to both data arrays describing what each image actually shows, and pointed thealtprop at it.Files:
src/components/sections/capabilities-grid.jsx,src/components/sections/tech-stack.jsx
Outstanding issues (6)
Moderate`/seo-audit` is an orphan page — indexable and in the sitemap, but linked from nowhere
src/components/layout/navbar.jsx · src/components/layout/footer.jsx
No internal link points at it. Orphan pages are crawled reluctantly and pass no internal link equity. There is also a judgement call underneath this: the page is indexable and candidly critical of the site it sits on, so it will surface for brand queries.
Suggested fix: Decide the intent. If it is a credibility asset, link it from the footer so it is crawlable and accumulates equity. If it is an internal working document, mark it
noindexand drop it from the sitemap. It should not stay in the current middle state. Left alone because the choice is yours, not a technical default.Why it is still open: Still open: both fixes are one line, but they point in opposite directions and the choice is a positioning decision about whether a public audit of your own site helps or hurts — not something to guess at.
MinorFigma profile still absent from `sameAs`
index.html — Person node
The audit lists
figma.com/@manan44, but the URL returns 403 to automated requests, which is Figma bot-blocking rather than proof the handle is wrong. Adding an unverifiable URL tosameAsis worse than omitting it: a broken entity link weakens the whole array.Suggested fix: Open the profile in a browser to confirm the handle, then append it to the
sameAsarray alongside Behance and Cal.com.Why it is still open: Still open: Figma serves 403 to every automated request, so the handle cannot be confirmed from here, and an unverified URL in
sameAsweakens the whole array.MinorLargest image asset is 262 kB
public/images/performance/development-workflow.webp
Already WebP and lazy-loaded below the fold, so the impact is small — but it is roughly a fifth of the site’s entire 1.3 MB image payload for one decorative illustration.
Suggested fix: Re-encode at a lower quality target or narrower intrinsic width; the existing
pnpm imagespipeline already does this work.Why it is still open: Still open, and the original fix note turns out to be wrong: the file is already at the width
optimize-images.mjstargets forperformance/(1200px) and is already WebP, so re-running the pipeline skips it. Squeezing it further means re-encoding a lossy WebP into another lossy WebP — visible quality loss for ~100 kB on a lazy-loaded decorative image. The real fix is a higher-quality source to re-encode from.MinorCase-study schema has no `datePublished` or `dateModified`
src/pages/case-study.jsx
The
CreativeWorkandWebApplicationnodes carry no dates, so nothing signals freshness or how recent the work is — a real consideration for a portfolio, where a prospect wants to know whether the last shipped project was last month or four years ago.Suggested fix: Add
dateModifiedfrom the same git-derived date the sitemap now uses, anddatePublishedper project once the real dates are known.Why it is still open: Still open:
dateModifiedis now derivable, butdatePublishedis not — I do not know when these projects actually shipped, and inventing a plausible date in structured data is a fabricated fact rather than a missing one. Needs the real dates from you.Minor`manifest.json` has no 192×192 or 512×512 icon
public/manifest.json
It declares 32×32, 180×180 and an SVG. Installability prompts and the Android home-screen icon want a 192 and a 512 PNG. This is a PWA and app-shell concern rather than a ranking factor, which is why it is Minor.
Suggested fix: Generate the two PNG sizes from
logo.svgand add them to theiconsarray.Why it is still open: Still open: it needs two new binary assets generated and committed, which is an asset change rather than the surgical metadata edit this pass was scoped to.
Moderate`react-helmet-async` is inert under React 19 and now sits in the dependency tree doing nothing
src/components/common/seo.jsx · src/main.jsx · src/entry-server.jsx
Its SSR context never populates, so the only thing
<Helmet>still does is pass children through to React 19’s native metadata handling. The app works, but it carries a dependency and a provider that imply behaviour they no longer deliver, and the build-time hoist exists to paper over the gap. Anyone readingseo.jsxwould reasonably assume Helmet is doing the work.Suggested fix: Drop
react-helmet-async, render the tags directly from<Seo>(React 19 hoists them natively on the client, and the prerenderer already handles the server side), and remove bothHelmetProviderwrappers.Why it is still open: Still open: it touches the SSR entry, the client entry and every page’s head component at once, which is a refactor rather than the surgical edit this request was scoped to. Behaviour is correct as it stands.
Not yet reviewed (5)
Field Core Web Vitals (LCP, INP, CLS on real devices)
Requires a browser and either a Lighthouse run or Chrome UX Report data. This pass checked the inputs that drive those numbers — asset weight, lazy-loading, bundle composition, prerendered HTML — but not the resulting scores. The Arbiris aspect-ratio finding above is the one place where a CLS problem is visible from static analysis alone.
Unused JavaScript and CSS coverage
Accurate numbers need a runtime coverage profile from DevTools. Static inspection shows a 45 kB Tailwind stylesheet (already purged) and a 242 kB React vendor chunk split away from route code, both of which look healthy, but "looks healthy" is not a measurement.
Live production verification
Everything here was verified against the local
dist/build. The deployed site has not been re-fetched since these changes, so thetrailingSlashredirect and the header rules invercel.jsonare unconfirmed in production.Off-page signals — backlinks, rankings, search-console data
Needs external tooling and account access that this pass did not have. The existing off-page section of this audit still stands as the plan.
Content accuracy of the six case studies
The case-study copy was drafted from the project data and stack rather than from first-hand build notes. The technical-decision narratives are plausible and internally consistent, but only the author can confirm they are true — and unverifiable claims are an E-E-A-T liability, not an asset.
Let's Build Something Great
Ready to turn
your idea into reality?
Whether you need a high-converting landing page, a modern business website, or a scalable web application, I'll help you build a fast, beautiful, and results-driven digital product.