How to vibe-code an SEO tool without losing control of your LLM

All of us use LLMs day by day. Most of us use them at work. Many people use them closely.

Folks in tech — sure, you — use LLMs at twice the rate of the general population. Many people spend greater than a full day every week utilizing them — sure, me.

Even these of us who depend on LLMs often get pissed off once they don’t reply the best way we would like.

Right here’s tips on how to talk with LLMs once you’re vibe coding. The identical classes apply if you end up in drawn-out “conversations” with an LLM UI like ChatGPT whereas making an attempt to get actual work carried out.

Select your vibe-coding atmosphere

Vibe coding is constructing software program with AI assistants. You describe what you need, the mannequin generates the code, and also you determine whether or not it matches your intent.

That’s the thought. In apply, it’s typically messier.

The very first thing you’ll must determine is which code editor to work in. That is the place you’ll talk with the LLM, generate code, view it, and run it.

I’m an enormous fan of Cursor and extremely advocate it. I began on the free Interest plan, and that’s greater than sufficient for what we’re doing right here. 

Honest warning – it took me about two months to maneuver up two tiers and begin paying for the Professional+ account. As I discussed above, I’m firmly within the “over a day every week of LLM use” camp, and I’d welcome the corporate.

 A number of choices are:

• Cursor: That is the one I exploit, as do most vibe coders. It has an superior interface and is well custom-made.
• Windsurf: The principle various to Cursor. It will possibly run its personal terminal instructions and self-correct with out hand-holding.
• Google Antigravity: Not like Cursor, it strikes away from the file-tree view and focuses on letting you direct a fleet of brokers to construct and take a look at options autonomously.

In my screenshots, I’ll be utilizing Cursor, however the rules apply to any of them. They even apply once you’re merely speaking with LLMs in depth.

Why prompting alone isn’t sufficient

You would possibly marvel why you want a tutorial in any respect. You inform the LLM what you need, and it builds it, proper? Which will work for a meta description or a superhero web optimization picture of your self, but it surely received’t reduce it for something reasonably advanced — not to mention a device or agentic system spanning a number of recordsdata.

One key idea to grasp is the context window. That’s the quantity of content material an LLM can maintain in reminiscence. It’s sometimes cut up throughout enter and output tokens.

GPT-5.2 presents a 400,000-token context window, and Gemini 3 Professional is available in at 1 million. That’s roughly 50,000 strains of code or 1,500 pages of textual content.

The problem isn’t simply hitting the restrict, particularly with giant codebases. It’s that the extra content material you stuff into the window, the more serious fashions get at retrieving what’s inside it.

Consideration mechanisms are likely to favor the start and finish of the window, not the center. On the whole, the much less cluttered the window, the higher the mannequin can concentrate on what issues.

If you need a deeper dive into context home windows, Matt Pocock has a great YouTube video that explains it clearly. For now, it’s sufficient to grasp placement and the price of being verbose.

A number of different ideas:

• One staff, one dream. Break your mission into logical phases, as we’ll do under, and clear the LLM’s reminiscence between them.
• Do your personal analysis. You don’t must develop into an knowledgeable in each implementation element, however you must perceive the directional choices for a way your mission may very well be constructed. You’ll see why shortly.
• When troubleshooting, belief however confirm. Have the mannequin clarify what’s taking place, evaluation it rigorously, and double-check crucial particulars in one other browser window.

Dig deeper: How vibe coding is changing search marketing workflows

How do you create content material that seems prominently in an AI Overview? Reply the questions the overview solutions.

On this tutorial, we’ll construct a device that extracts questions from AI Overviews and shops them for later use. Whereas I hope you discover this use case precious, the true aim is to stroll via the phases of correctly vibe coding a system. This isn’t a shortcut to successful an AI Overview spot, although it might assist.

Step 1: Planning

Earlier than you open Cursor — or your device of selection — get clear on what you need to accomplish and what sources you’ll want. Assume via your strategy and what it’ll take to execute.

Whereas I famous to not launch Cursor but, it is a high-quality time to make use of a standard search engine or a generative AI.

I have a tendency to start out with a easy sentence or two in Gemini or ChatGPT describing what I’m making an attempt to perform, together with a listing of the steps I feel the system would possibly must undergo. It’s OK to be incorrect right here. We’re not constructing something but.

For instance, on this case, I’d write:

I’m an web optimization, and I need to use the present AI Overviews displayed by Google to encourage the content material our authors will write. The aim is to extract the implied questions answered within the AI Overview. Steps would possibly embrace:

1 – Choose a question you need to rank for.
2 – Conduct a search and extract the AI Overview.
3 – Use an LLM to extract the implied questions answered within the AI Overview.
4 – Write the inquiries to a saveable location.

With this in hand, you’ll be able to head to your LLM of selection. I want Gemini for UI chats, however any fashionable mannequin with strong reasoning capabilities ought to work.

Begin a brand new chat. Let the system know you’ll be constructing a mission in Cursor and need to brainstorm concepts. Then paste within the planning immediate.

The system will instantly present suggestions, however not all of it is going to be good or in scope. For instance, one response advised monitoring the AI Overview over time and operating it in its personal UI. That’s past what we’re doing right here, although it might be value noting.

It’s additionally value noting that fashions don’t at all times counsel the only path. In a single case, it proposed a posh methodology for extracting AI Overviews that might probably set off Google’s bot detection. That is the place we return to the listing we created above.

Step 1 might be straightforward. We simply want a discipline to enter key phrases.

Step 2 might use some refinement. What’s probably the most easy and dependable method to seize the content material in an AI Overview? Let’s ask Gemini.

I’m already accustomed to these providers and continuously use SerpAPI, so I’ll select that one for this mission. The primary time I did this, I reviewed choices, in contrast pricing, and requested just a few friends. Making the incorrect selection early could be expensive.

Step 3 additionally wants a better look. Which LLMs are greatest for query extraction?

That stated, I don’t belief an LLM blindly, and for good cause. In a single response, Claude 4.6 Opus, which had lately been launched, wasn’t even thought of.

After a few back-and-forth prompts, I instructed Gemini:

• “Now, be crucial of your options and the benchmarks you’ve chosen.”
• “The textual content might be quick, so value isn’t a difficulty.”

We then got here round to:

For this mission, we’re going with GPT-5.2, because you probably have API entry or, on the very least, an OpenAI account, which makes setup straightforward. Name it a hunch. I received’t add an LLM choose on this tutorial, however in the true world, I strongly advocate it.

Now that we’ve carried out the back-and-forth, we now have extra readability on what we’d like. Let’s refine the define:

I’m an web optimization, and I need to use the present AI Overviews displayed by Google to encourage the content material our authors will write. The concept is to extract the implied questions answered within the AI Overview. Steps would possibly embrace:

1 – Choose a question you need to rank for.
2 – Conduct a search and extract the AI Overview utilizing SerpAPI.
3 – Use GPT-5.2 Pondering to extract the implied questions answered within the AI Overview.
4 – Write the question, AI Overview, and inquiries to W&B Weave.

Earlier than we transfer on, be sure to have entry to the three providers you’ll want for this:

• SerpAPI: The free plan will work.
• OpenAI API: You’ll must pay for this one, however $5 will go a great distance for this use case. Assume months. 
• Weights & Biases: The free plan will work. (Disclosure: I’m the top of web optimization at Weights & Biases.)

Now let’s transfer on to Cursor. I’ll assume you’ve gotten it put in and a mission arrange. It’s fast, straightforward, and free. 

The screenshots that observe mirror my most popular structure in Editor Mode.

Step 2: Set the groundwork

If you happen to haven’t used Cursor earlier than, you’re in for a deal with. Certainly one of its strengths is entry to a spread of fashions. You’ll be able to select the one that matches your wants or decide the “greatest” possibility based mostly on leaderboards.

I are likely to gravitate towards Gemini 3 Professional and Claude 4.6 Opus.

If you happen to don’t have entry to all of them, you’ll be able to choose the non-thinking fashions for this mission. We additionally need to begin in Plan mode.

Let’s start with the mission immediate we outlined above.

Observe: You might be requested whether or not you need to permit Cursor to run queries in your behalf. You’ll need to permit that.

Now it’s time to shuttle to refine the plan that the mannequin developed from our preliminary immediate. As a result of it is a pretty easy activity, you would possibly assume we might leap straight into constructing it, which might be unhealthy for the tutorial and in apply. If you happen to thought that, you’d be incorrect. People like me don’t at all times talk clearly or totally convey our intent. This strategy planning stage is the place we make clear that.

After I enter the directions into the Cursor chat in Planning mode, utilizing Sonnet 4.5, it kicks off a dialogue. One of many nice issues about this stage is that the mannequin typically surfaces angles I hadn’t thought of on the outset. Beneath are my replies, the place I reply every query with the relevant letter. You’ll be able to add context after the letter if wanted.

An instance of the mannequin suggesting angles I hadn’t thought of seems in query 4 above. It might be useful to move alongside the context snippets. I opted for B on this case. There are apparent circumstances for C, however for velocity and token effectivity, I retrieve as little as potential. Intent and associated concerns are exterior the scope of this text and would add complexity, as they’d require a choose.

The system will output a plan. Learn it rigorously, as you’ll virtually actually catch points in the way it interpreted your directions. Right here’s one instance.

I’m instructed there is no such thing as a GPT-5.2 Pondering. There may be, and it’s famous within the announcement. I’ve the system double-check just a few particulars I need to verify, however in any other case, the plan appears to be like good. Claude additionally famous the format the system will output to the display screen, which is a pleasant contact and one thing I hadn’t specified. That’s what companions are for.

Lastly, I at all times ask the mannequin to assume via edge circumstances the place the system would possibly fail. I did, and it returned a listing. From that listing, I chosen the circumstances I wished addressed. Others, like what to do if an AI Overview exceeds the context window, are so unlikely that I didn’t trouble.

A number of closing tweaks addressed these objects, together with one I added myself: what occurs if there is no such thing as a AI Overview?

I’ve to provide credit score to Tarun Jain, whom I discussed above, for this subsequent step. I used to repeat the define manually, however he advised merely asking the mannequin to generate a file with the plan. So let’s direct it to create a markdown file, plan.md, with the next instruction:

Construct a plan.md together with the reviewed plan and plan of motion for the implementation. 

Bear in mind the context window problem I mentioned above? If you happen to begin constructing out of your present state in Cursor, the preliminary directives could find yourself in the midst of the window, the place they’re least accessible, since your mission brainstorming occupies the start.

To get round this, as soon as the file is full, evaluation it and ensure it precisely displays what you’ve brainstormed.

Step 3: Constructing

Now we get to construct. Begin a brand new chat by clicking the + within the prime proper nook. This opens a brand new context window.

This time, we’ll work in Agent mode, and I’m going with Gemini 3 Professional.

Arguably, Claude 4.6 Opus is likely to be a technically more sensible choice, however I discover I get extra correct responses from Gemini based mostly on how I talk. I work with far smarter builders preferring Claude and GPT. I’m undecided whether or not I naturally talk in a approach that works higher with Gemini or if Google has skilled me over time.

First, inform the system to load the plan. It instantly begins constructing the system, and as you’ll see, it’s possible you’ll must approve sure steps, so don’t step away simply but.

As soon as it’s carried out, there are solely a few steps left, hopefully. Fortunately, it tells you what they’re.

First, set up the required libraries. These embrace the packages wanted to run SerpAPI, GPT, Weights & Biases, and others. The system has created a necessities.txt file, so you’ll be able to set up all the pieces in a single line.

Observe: It’s greatest to create a digital atmosphere. Consider this as a container for the mission, so downloaded dependencies don’t combine with these from different initiatives. This solely issues in case you plan to run a number of initiatives, but it surely’s easy to arrange, so it’s value doing.

Open a terminal:

Then enter the next strains, separately:

• python3 -m venv .venv
• supply .venv/bin/activate
• pip set up -r necessities.txt

You’re creating the atmosphere, activating it, and putting in the dependencies inside it. Maintain the second command useful, because you’ll want it any time you reopen Cursor and need to run this mission.

You’ll know you’re within the right atmosphere once you see (.venv) at the start of the terminal immediate.

Once you run the necessities.txt set up, you’ll see the packages load.

Subsequent, rename the .env.instance file to .env and fill within the variables.

The system can’t create a .env file, and it received’t be included in GitHub uploads in case you go that route, which I did and linked above. It’s a hidden file used to retailer your API keys and associated credentials, which means info you don’t need publicly uncovered. By default, mine appears to be like like this.

I’ll fill in my API keys, sorry, can’t present that display screen, after which all that’s left is to run the script.

To do this, enter this within the terminal:

python most important.py "your search question"

If you happen to neglect the command, you’ll be able to at all times ask Cursor.

Oh no … there’s an issue!

I’m constructing this as we go, so I can present you tips on how to deal with hiccups. After I ran it, I hit a crucial one.

It’s not discovering an AI Overview, although the phrase I entered clearly generates one.

Fortunately, I’ve a wide-open context window, so I can paste:

• A picture displaying that the output is clearly incorrect.
• The code output illustrates what the system is discovering.
• A hyperlink (or generally merely textual content) with extra info to direct the answer. 

Fortuitously, it’s straightforward so as to add terminal output to the chat. Choose all the pieces out of your command via the complete error message, then click on “Add to Chat.”

It’s essential to not rely solely on LLMs to search out the knowledge you want. A fast search took me to the AI Overview documentation from SerpAPI, which I included in my follow-up directions to the mannequin.

My troubleshooting remark appears to be like like this.

Discover I inform Cursor to not make adjustments till I give the go-ahead. We don’t need to refill the context window or practice the mannequin to imagine its job is to make errors and take a look at fixes in a loop. We cut back that danger by reviewing the strategy earlier than modifying recordsdata.

Glad I did. I had a hunch it wasn’t retrieving the code blocks correctly, so I added one to the chat for added evaluation. Remember the fact that LLMs and bots could not see all the pieces you see in a browser. If one thing is essential, paste it in for example.

Now it’s time to strive once more.

Wonderful, it’s working as we hoped.

Now we now have a listing of all of the implied questions, together with the consequence chunks that reply them.

Dig deeper: Inspiring examples of responsible and realistic vibe coding for SEO

Logging and tracing your outputs

It’s a bit messy to rely solely on terminal output, and it isn’t saved when you shut the session. That’s what I’m utilizing Weave to handle.

Weave is, amongst different issues, a device for logging immediate inputs and outputs. It provides us a everlasting place to evaluation our queries and extracted questions. On the backside of the terminal output, you’ll discover a hyperlink to Weave.

There are two traces to observe. The primary is what this was all about: the analyze_query hint.

Within the inputs, you’ll be able to see the question and mannequin used. Within the outputs, you’ll discover the complete AI Overview, together with all of the extracted questions and the content material every query got here from. You’ll be able to view the complete hint here, in case you’re .

Now, once we’re writing an article and need to ensure we’re answering the questions implied by the AI Overview, we now have one thing concrete to reference.

The second hint logs the immediate despatched to GPT-5.2 and the response.

This is a vital a part of the continuing course of. Right here you’ll be able to simply evaluation the precise immediate despatched to GPT-5.2 with out digging via the code. If you happen to begin noticing points within the extracted questions, you’ll be able to hint the issue again to the immediate and get again to vibing along with your new good friend, Cursor.

Construction beats vibes

I’ve been vibe coding for a few years, and my strategy has developed. It will get extra concerned after I’m constructing multi-agent programs, however the fundamentals above are at all times in place.

It might really feel quicker to drop a line or two into Cursor or ChatGPT. Strive that just a few occasions, and also you’ll see the selection: hand over on vibe coding — or be taught to do it with construction.

Maintain the vibes good, my associates.