Should I vibe code
WordPress SEO guidance, internal linking, redirects, schema, and support
You will edit sentences a human wrote until a threshold you invented last night turns green.
?
Their verdict, the Premium price and the build-time estimate come from their entry, MIT-licensed. Checked 2026-08-04.
?
Our verdict, the regret score and everything below it. Editorial and unsponsored — nobody can pay to be moved.
The honest answer
why the verdict is what it is
The free plugin already emits the tags, so what you would be rebuilding is the Premium half — and Premium is two features that fail in opposite directions. One is advice: keyphrase distribution, readability, the traffic light. The difficulty there is not the code, it is that you will obey it. A generated scorer invents its thresholds, then presents them as a measurement, and you will edit sentences a human wrote for other humans until your own made-up number goes green. Nothing errors, no test fails, and the damage is to the writing. The other feature is the redirect manager, which is not advice but state, wired into WordPress events so that trashing a post offers to 301 it somewhere. That table is the only thing keeping a decade of inbound links pointed at pages that still exist, and a 301 is the most durable thing a website can say: browsers cache it aggressively, Google reads it as a promise, and you cannot reach into anybody's cache to withdraw it. Get one rule's trailing slash backwards and every returning visitor loops until their own cache expires. Sitting under both is what the other two WordPress entries here already argue — a plugin is not a project you finish, it is PHP you adopt inside a CMS that ships two majors a year — and this one adds server-config writes and live edits to published content on top.
What actually breaks
not "if". the specific failures.
- The redirect table, where one rule with the trailing slash on the wrong side becomes a loop that every returning visitor is now caching for you
- Automatic redirects on deletion, which fire once per row during a bulk trash and rewrite forty URLs from a single click with no undo
- 301s you have already served, which outlive the fix in browser caches and Google's index by weeks
- The readability and keyphrase scorer, a pile of invented cut-offs wearing the costume of a measurement
- Internal-linking suggestions that write into published post content, where a bad match is a live edit to a page somebody is reading
- The .htaccess or nginx rules file, if redirects go there, because your plugin is now editing server configuration at runtime
- Admin AJAX and REST routes shipped without a nonce and a capability check, which is the standard WordPress plugin finding and is scanned for continuously
- The 404 log, if you add one, which is an unbounded table fed by every bot on the internet
- The treadmill you joined: WordPress majors, PHP deprecations, block-editor churn, and Google changing its mind about what a canonical means
Is that you?
the verdict is a default, not a law
- It emits metadata and nothing else, on one site you own — which is the free plugin's job, done by you for fun
- Redirects, if they exist at all, are generated as a rules file you read before deploying it yourself
- Nothing it does writes into post content
- You watch Search Console often enough to notice a cliff in days rather than months
- No other site will ever run it
- It creates redirects automatically in response to content edits
- The site has years of inbound links, which makes the redirect table an archive rather than a config file
- The scoring UI will be shown to writers who will believe it
- A suggestion feature can modify published posts without a human approving a diff
- You are handing it to a client whose traffic is their income
If you build it anyway
the checklist, then the prompt that enforces it
- Read the free plugin first. Yoast SEO Free is GPL and on GitHub, and it already does the entire metadata layer — the honest question is which Premium feature you actually want.
- Redirects are a separate project from metadata. Ship the tag emitter, run it for a month, then decide whether you still want to own a redirect table.
- Never create a redirect implicitly. No "we noticed you deleted a post" automation without a human confirming each individual rule.
- Validate at write time: reject loops, cap chains at one hop, normalise trailing slashes and case on both sides, and refuse a rule whose source still resolves to a live page.
- Default to 302 until you are certain. A temporary redirect is not cached the way a 301 is, so a mistake costs you an afternoon instead of a month.
- Keep an append-only audit log of every redirect change, and an export to a plain nginx or Apache rules file, because the redirects need to be able to outlive the plugin.
- If you build a content scorer, keep every threshold in one documented file and label the output as a rule of thumb in the UI. "Readability: 62" invites obedience; "heuristic" does not.
- A suggestion feature suggests. Show a diff, require a click, write a revision, and never touch post_content on a cron.
- Nonce check and current_user_can on the first line of every admin endpoint, before the handler body exists.
I am writing my own version of a premium WordPress SEO plugin for one site I own.
The two failures I care about are a redirect rule that loops or points somewhere
wrong — permanently, because 301s are cached — and an advice UI confident enough
that I rewrite good prose to satisfy it. Order the work accordingly.
1. Phase one emits metadata only: title, description, canonical, robots, OpenGraph,
JSON-LD. No redirects, no writes to post content. Ship that and stop.
2. Add integration tests that render a post, a page, an archive, a paginated archive
and a 404, asserting the exact robots directive. Keep them green forever.
3. Before the first redirect can be created, build its export: a command that dumps
the whole table as nginx and Apache rules. If it cannot leave, do not fill it.
4. Redirect rules are validated at write time — no loops, chain length capped at one
hop, source and destination normalised for case and trailing slash — and every
change appends to an audit log with user and timestamp.
5. New rules default to 302. Promoting a rule to 301 is a separate, explicit action,
and the UI says out loud that a 301 is cached in browsers I will never touch again.
6. Never create a redirect as a side effect of deleting or renaming a post. Offer it,
show me the exact rule, and require one confirmation per rule. Bulk operations do
not get a bulk redirect.
7. Do not write to .htaccess or any server config file at runtime. Store rules in the
database and serve them from a template_redirect hook.
8. If I ask for internal-linking suggestions, they suggest only: render a diff,
require a click, save a revision. Nothing edits published content unattended.
9. Any content scoring keeps its thresholds in one file with comments explaining
where each number came from, and the UI labels them as heuristics. If I ask for a
single "SEO score" number, tell me why that is a bad idea first.
10. Every admin-ajax and REST route begins with a nonce check and current_user_can,
and deactivation restores the theme's original head output byte for byte.
11. Out of scope unless I ask again: Search Console OAuth, 404 logging, and anything
that calls an AI model on save. Finish by reminding me Yoast Free and the
open-source Redirection plugin already cover most of this for nothing.That one keeps you out of trouble. For the prompt that actually builds it, canivibecodeit.com has one.
their build prompt ↗Or don’t build it
the boring option, and the way back out
The site is how you get paid. Premium is $118.80 a year for one site, under ten dollars a month, and it comes with somebody else's regression suite over the redirect manager — which is the feature whose bugs are cached in strangers' browsers. But the first honest answer is not "buy it": Yoast Free already does titles, meta, canonicals and sitemaps, and the open-source Redirection plugin does redirects with a 404 log. Pay for Premium when you want the redirect manager, the linking suggestions and someone to email at 9am, not because your blog needs the traffic light.
$9.9/mo is cheaper than your weekend.
Two halves, two exits. Metadata is recomputable, so keep every value in post meta and wp_options under one prefix — and if you name the keys the way Yoast names its own (_yoast_wpseo_*), the real plugin will import your work in one click the day you give up. Redirects are the opposite: they are live state with other people's links pointing at them, so keep an export that writes the whole table as nginx or Apache rules and move those to the server before you deactivate anything. A redirect that only exists inside a plugin you have stopped maintaining is a time bomb with a WordPress update as its fuse.
The GPL version of the exact product on this page, and it already ships the whole metadata and sitemap layer.
Open-source WordPress redirect manager with 404 logging, and a good read before hand-rolling a redirect table.
A lighter open-source WordPress SEO plugin, small enough to read end to end in an evening.
Questions
Two other WordPress SEO plugins here are already DEMO ONLY. What is different about this one?
The feature set you would be cloning. AIOSEO's entry is about emitting metadata and the silent noindex; Rank Math's is about pulling Search Console data into wp-admin and the OAuth token that ends up in wp_options. Yoast's paid half is neither: it is a redirect manager wired to content events, plus an advice engine. So this page is about permanence — cached 301s and edits to published posts — rather than about output or credentials. Same band, different failure.
Is the redirect manager really the dangerous part?
It is the part you cannot take back. A wrong meta tag is fixed by a deploy and re-crawled eventually. A wrong 301 is cached by every browser that saw it, honoured by intermediaries, and treated by search engines as a statement that the old URL is gone for good. Serve a loop and returning visitors keep looping after the fix, because their cache is not yours to clear. That is why the guardrails say default to 302 and require one human confirmation per rule.
canivibecodeit says YES, one sitting. Why harsher?
Because they are answering whether an agent can produce a working plugin, and it can — the metadata emitter really is one sitting. We are answering whether you should run it for four years. The parts of Premium worth paying for are precisely the parts whose mistakes are cached, published or written into other people's content, and the WordPress upgrade treadmill is a subscription you pay in weekends rather than dollars.
Every week, someone ships something they shouldn’t have.
New verdicts, the worst thing that landed in the trap, and the occasional incident report. No other email, ever.
One inverted conditional and your SEO plugin ships noindex. Nothing errors. You find out in five weeks.
Your wp_options table now holds a live Google refresh token, next to a 404 log nobody put a ceiling on.
Crawling your own site and reading the headers is a script, not a licence.
last reviewed 2026-08-05 · verdict is editorial and unsponsored · shared entry data from canivibecodeit under MIT · not legal advice