Should I vibe code
SEO content optimization and AI writing based on SERP analysis
SERP analysis means scraping search results, which is a fragile foundation to build on.
?
Their verdict, the Essential/Standard price and the build-time estimate come from their entry, MIT-licensed. Checked 2026-08-03.
?
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 content scoring is a spreadsheet with opinions. The data acquisition is where this breaks, repeatedly, without warning.
What actually breaks
not "if". the specific failures.
- SERP data, which requires either scraping search results against their terms or paying for an API that already does
- Correlation presented as instruction, so 'pages that rank use this word 14 times' becomes a target rather than an observation
- Optimising to a ranking function that changes without notice and is not published
- Content written to hit a score, which reads like content written to hit a score
- Volume, because the tool makes producing thin pages cheap and search engines have spent a decade penalising exactly that
You publish thirty pages, each scoring above ninety in your own tool, each hitting the term frequencies the top results share. Six weeks later traffic to all thirty falls together. Nothing was penalised in a way you can see, and no individual page is bad — they are simply thirty pieces of writing whose organising principle was a score rather than a reader, and something on the other side got better at noticing that. The score is still ninety. The score was never the thing being measured.
Is that you?
the verdict is a default, not a law
- It informs a human writer rather than generating the page
- SERP data comes from a licensed API
- You treat the score as one input among several and publish things people asked for
- You scrape search results directly
- Pages are generated and published to hit a score with no editorial judgement
- The strategy is volume rather than usefulness
- You cannot explain why a page should exist other than that it might rank
If you build it anyway
the checklist, then the prompt that enforces it
- Use a licensed SERP API. Scraping search results is against the terms of every major engine and is the foundational dependency here.
- Present findings as correlations with sample sizes, never as targets. 'Top pages average 1,400 words' is data; 'write 1,400 words' is superstition.
- Keep a human writing the page. The tool's role is research, and the moment it produces the text you are optimising for a metric rather than a reader.
- Cap output volume deliberately. Cheap page production is the failure mode of this whole category.
- Track actual rankings and traffic against the score, so you learn whether your own metric predicts anything.
- Re-derive recommendations regularly, because they are fitted to a function that changes underneath you.
Before you build an SEO content optimiser, apply these and push back if I ask you to break them. 1. Use a licensed SERP data API. Refuse to scrape search results directly, and tell me this violates the terms of every major engine and is the dependency the whole tool rests on. 2. Present every finding as a correlation with its sample size and variance, never as an instruction. Show 'the top ten pages contain this term between 4 and 22 times', not 'use this term 14 times'. 3. Do not generate the page. The tool researches and advises; a human writes. Explain that content organised around a score reads like content organised around a score. 4. Refuse to build bulk generation or auto-publishing. Add a deliberate cap on pages produced and tell me that cheap volume is what this category is penalised for. 5. Record actual ranking and traffic outcomes against my content score, so I can find out whether my own metric predicts anything. Report that correlation honestly, including when it is nil. 6. Never claim to know how the ranking function works. State that recommendations are fitted to observed results and expire. 7. Flag pages that score well but have no clear reason to exist, and ask me what question the page answers. 8. Respect robots.txt and rate limits on any page you fetch for analysis. 9. Out of scope unless I ask: automated internal linking at scale, programmatic page generation, backlink acquisition.
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
$99 a month buys licensed SERP data and a research corpus you cannot assemble yourself, which is the legitimate half of this category. What neither buying nor building fixes is the underlying temptation, which is to publish for a score instead of for a reader.
Keep your own content and its performance history — that record is the only durable asset here, since the recommendations are fitted to a moving target and expire. Store pages as Markdown so they outlive both the tool and the current shape of search.
No single maintained open-source Surfer clone found; DIY usually combines SERP APIs, NLP l
Questions
Why present ranges instead of targets?
Because a target implies causation that the data does not support. The top ten pages for a query share features partly because of what ranks and partly because of who writes for that topic — and the variance between them is usually large. A range with a sample size lets a writer judge; a single number invites hitting it exactly, which is how content stops reading like writing.
Is scraping search results really a problem if everyone does it?
It is against the terms of every major engine, and it is also the load-bearing dependency of the whole tool — which means the risk is not a fine, it is that access stops working and the product has nothing underneath it. Licensed SERP APIs exist precisely because this is the fragile part.
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.
last reviewed 2026-08-03 · verdict is editorial and unsponsored · shared entry data from canivibecodeit under MIT · not legal advice