Should I vibe code
Build a content brief and compare a draft against transparent terms from selected pages
The cheapest SERP API is $25 a month. NeuronWriter is $23. You cannot buy the ingredients for less than the meal.
?
Their verdict, the Bronze 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 code here is not the obstacle, and pretending otherwise wastes your weekend on the wrong problem. Fetch a SERP, pull the text off the pages it returns, count terms, diff that against your draft, render a brief — an agent will hand you all of it before dinner, and NeuronWriter is unusually honest about the method being that simple. The obstacle is the input. Every analysis begins with a search result page you are not allowed to fetch automatically, which means you either scrape Google until your IP stops working or you buy the data, and the arithmetic of buying is brutal: NeuronWriter's entry tier is $23 a month, and the entry plan at the best-known SERP API is $25. You cannot buy the ingredients for less than the finished meal. Build it to understand what the score is made of — that is genuinely worth an evening — then notice you are now maintaining a data pipeline in order to save nothing.
What actually breaks
not "if". the specific failures.
- SERP acquisition, first and repeatedly — automated search queries are against Google's terms, and the practical enforcement is not a lawyer but a captcha that arrives on the day you had forty briefs queued
- The competitor crawl, because the pages you need the text from are increasingly rendered client-side, paywalled or behind a bot check, and a page that returns nothing quietly lowers every term count
- Term extraction, which drifts as the corpus drifts — the same keyword analysed in March and in September gives different targets, and nothing in your tool records which corpus produced which brief
- The score, when a page hits 84 because it resembles four pages you selected that were answering a different question entirely
- Cost modelling, once you notice the LLM tokens are the cheap part and the SERP calls are the bill
- The AI-visibility feature, if you try to copy it — tracking whether your page gets cited inside an assistant's answer means querying assistants on a schedule and parsing citations, which is a harder acquisition problem than the SERP was
Is that you?
the verdict is a default, not a law
- You want to see what a content score is actually made of, and the answer is going to be 'term frequency, weighted'
- You run a handful of briefs a month and already pay for SERP data for something else
- Every brief stores the exact competitor set and fetch date that produced it
- The output is a checklist a human reads, not a number a freelancer is paid against
- The plan is to query Google directly on a schedule and hope
- You will publish at volume against a target your own arithmetic invented
- Somebody else — a writer, an agency, a contractor — gets paid or graded on the score your code produces
- You need this to keep working while you are on holiday, because SERP scraping breaks on days you did not choose
If you build it anyway
the checklist, then the prompt that enforces it
- Decide the data source before you write a line. Either you are paying a SERP vendor, in which case price it honestly against the $23 subscription, or you are scraping Google, in which case say so out loud so it is a decision rather than a drift.
- Store the corpus, not just the score. Every brief should carry the exact URLs, the fetch timestamp and the raw extracted text, because a target you cannot reproduce is a target you cannot argue with in three months.
- Make a failed page fetch loud. A competitor page that returns a bot-check body extracts to almost nothing and silently drags every term count down; a missing page should abort the analysis, not average into it.
- Never render the score as a single number without the term list beside it. The number is an opinion; the terms are the evidence, and NeuronWriter shows both for a reason.
- Cap the target. A recommendation of 'use this term 14 times' produced by term-frequency arithmetic is how pages end up unreadable — clamp anything the model suggests to a range a person would actually write.
- Keep the brief and the draft in separate files under version control, so you can see which brief produced which page and whether either of them helped.
I am building a content brief and on-page optimisation tool. Apply these before any scraping or scoring code exists, and argue with me when I cross one. 1. Ask me first where SERP data comes from. If the answer is 'scrape Google', tell me that is against their terms, that the practical failure is a captcha on a day I am busy, and quote me the cost of a real SERP API before writing the fetcher. 2. Build storage before analysis. Each run persists the query, the ranked URLs, the fetch timestamp and the raw extracted text of every page. A brief whose corpus I cannot reproduce is worthless three months later. 3. Treat a failed or bot-blocked page fetch as an error that aborts the run. Never let a page that extracted 40 characters average quietly into the term counts. 4. Extract page text with a real readability pass, and record how many characters each page contributed. Show me that table. 5. Present the term list before the score, always. If I ask for a single optimisation number, add it, but never render it without the evidence beside it and never imply it is calibrated against ranking outcomes — it is not. 6. Clamp every frequency recommendation to a range a human would plausibly write, and refuse to emit instructions like 'use this phrase 22 times'. 7. Rate-limit and cache aggressively. The same keyword analysed twice in a week should reuse the stored corpus rather than re-buying it. 8. Keep drafting separate from scoring. Do not build a loop where a model writes text to satisfy a score my own code invented — that is how forty unreadable pages get published. 9. Out of scope: rank tracking, plagiarism detection, and AI-answer citation monitoring. Each is a separate data-acquisition project. Say so, and tell me NeuronWriter's entry tier is $23 a month.
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
You publish more than a couple of pieces a month, which is roughly the point where SERP acquisition stops being a curiosity and starts being a chore with a bill. Twenty-three dollars covers the data, the crawling, the term extraction and — the part nobody self-builds — somebody else watching for the day Google changes shape. If you are billing a client for optimised content, buy it and spend the weekend on the content.
$23/mo is cheaper than your weekend.
The briefs are the artefact worth keeping, so write each one to a markdown file with its keyword, its competitor URLs, its fetch date and its term table embedded at the bottom. Do that from run one and the tool is disposable: you can drop the pipeline, subscribe to NeuronWriter or Surfer, and still be able to answer why a page was written the way it was.
Open-source technical SEO auditing application.
Self-hosted rank tracker, and a working example of what SERP data costs once you have to source it yourself.
Questions
How is this different from the Frase and Surfer verdicts?
Frase's entry is about the legitimacy of the score — you can build a scoreboard but not a referee. Surfer's is about fragility in the acquisition layer. NeuronWriter's is arithmetic. It is the cheapest serious tool in the category, its method is the most transparent, and that combination means the interesting question is not whether you can write the code but whether the input costs less than the subscription. It does not.
Can I not just use a cheaper wholesale SERP feed?
You can, and they are cheaper per query. They also come with minimum top-ups, their own terms about what you may do with the results, and no relationship with you when a result set goes weird. Once you have added a monthly floor, an LLM bill and your own time to the ledger, the $23 stops looking like the expensive option and starts looking like the price of not thinking about any of it.
Is the optimisation score meaningful at all?
It is meaningful in the way a checklist is: it catches the case where you wrote two thousand words about a topic and never mentioned the thing everyone ranking mentions. It is not a ranking prediction, and NeuronWriter is more candid about that than most. The danger with a homemade version is not that the number is wrong, it is that it is yours, so nobody in the room will argue with it.
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.
SERP analysis means scraping search results, which is a fragile foundation to build on.
You can build the scoreboard. You cannot build the referee — and the score is the thing people obey.
A recommendation to use the phrase eleven more times is keyword stuffing with a progress bar on it.
last reviewed 2026-08-05 · verdict is editorial and unsponsored · shared entry data from canivibecodeit under MIT · not legal advice