People assume that real businesses get built by other people with more resources. It's how great ideas get stuck in notes apps and conversations.

But you don't need six months and a developer's salary to validate your concept. You just need a weekend. Describe what you want to exist and Lovable builds it: a real site, a real product, ready to accept real payments from real customers. Not a mockup to show people. An actual thing they can use.

You don't need anyone's sign-off to get started. Take $5 off your first month.

Find out this weekend

_{^{Discount of $5 applies to first month of any paid plan. You need to sign up via this webpage in order to claim the discount. The discount applies to new and returning users without an active paid plan. Terms apply.}}

Most of the reports I send clients start life as a quick email. One landed on Friday from a client's marketing team, asking for something that sounded simple: a list of their URLs mapped to the prompts and markets they want to win in AI search.

The team wants it so they can approach the third-party sites that AI answers already trust, and get the brand mentioned and linked the way its competitors are. They were clear they didn't need a detailed report, just the URLs. It's a good ask, and it's exactly the kind of job that used to mean an afternoon of exports, VLOOKUPs and a spreadsheet nobody enjoys reading.

This week I built it a different way, by connecting Obsero to Claude through our MCP connector. The URL list was the quick part, so I went a step further and asked Claude to turn it into a hit list the PR team could work from, shared as a link. Below is the step-by-step, using Levi's as the worked example rather than the client, including how to set it up so your clients' data stays where it should.

MCP (Model Context Protocol) is the standard that lets Claude plug into other tools. Once Obsero is connected, Claude can query your AI search data directly while you talk to it, so there's nothing to export or copy across, and no code to write.

It's written for the solo consultant or small agency serving a handful of clients, although it works just as well in-house.

In the Claude app, go to Settings, then Connectors, and click "Add custom connector". Name it Obsero and paste in the MCP URL, https://mcp.obsero.ai. Claude sends you to Obsero to sign in, where you click "Allow access" and then head back to Claude. To check it's working, ask Claude whether it's connected to the Obsero MCP. Our setup guide walks through each screen.

Once it's connected, Claude can see Obsero's MCP tools, such as visibility, citations and cited pages, and uses them whenever you ask a question. You sign in through Obsero itself, so there are no keys or passwords to copy around, and nothing sensitive ends up sitting in a chat, a shared doc or a Slack message.

If you're on a Team or Enterprise plan, only an owner can add connectors for the organisation (under Admin settings, then Connectors), and each member then connects it and switches it on for themselves.

Each connection covers one Obsero organisation, so if you look after several clients the cleanest setup is one connection per client, named so you can't mix them up. Only switch on the connection for the client you're working on in that chat.

Before you put client data anywhere near an AI model, it's worth knowing what happens to it, and there are a few settings that matter here.

With the connector switched on in a new chat, you can ask for the report in plain English and Claude calls Obsero's MCP tools to fetch the data it needs. This is roughly the prompt I used for Levi's:

Two things make the output better. Ask for one market at a time and then several together because, as the client spotted, one market on its own can show only a handful of pages, while adding another often reveals more. Then ask Claude how much of the data it's looking at, so you know whether a list is complete or just the top 50 of several hundred.

Claude builds the report as an artifact, which is a web page that stays private to you until you share it. I check every URL in it before anything goes to a client, then share the link rather than attaching a file.

To keep it current, ask Claude to turn it into a scheduled task, something like "every Monday at 8am, re-run this through the Obsero connector and update the Levi's report". Claude refreshes the same page on that schedule, so the client always sees the latest version at the same link, and I can check what's moved before they open it.

The Levi's pages tab is where the story for the PR team starts. Switch it to prompt visibility and filter by how often Levi's appears, and you get a shortlist of the prompts it's losing. Here I've filtered to 20% to 60%, the middle ground where Levi's shows up but isn't winning, and the pattern is clear: plus size, petite and body-type questions. Those are the prompts to pitch around, because that's where a handful of extra mentions has the best chance of moving the number.

Here's what the report lets you and the PR team do with it:

A year ago this kind of report needed a data person, a BI tool or a long weekend. Now a one-person consultancy can turn a client email into something that looks like it came from a much bigger agency, while the thinking stays with the person who actually knows the client.

The report is only as good as the prompts you track and the person checking it, though, so let Claude do the assembly and keep the judgement for yourself.

If your analytics showed a dip in traffic from ChatGPT this week, check it again before you panic. After the GPT-6 rollout, the links inside ChatGPT's answers stopped carrying the utm_source=chatgpt.com tag, and because most ChatGPT clicks arrive without a referrer, tools like GA4 had nothing to go on and filed those visits as Direct. Search Engine Journal covered it on 15 October, reporting that OpenAI had confirmed a bug and started rolling out a fix.

We can pinpoint when it happened, because Obsero stores the citation URLs ChatGPT puts in its answers every day.

Two details are worth knowing. The model didn't roll back, so the tag is returning on GPT-6 itself. The links in the sources list beneath the answer haven't recovered yet, though, and have sat at around 47% tagged since the change, down from about 65% before it.

So if your ChatGPT traffic dipped from 9 October, it's most likely a labelling problem rather than people suddenly visiting less, and it's worth checking whether your Direct traffic rose over the same days. Our data is what ChatGPT put in its answers rather than visits to your site, so it shows when the tag went missing, not how much traffic you lost.

This is a good example of how little the big AI platforms give us. There's no Search Console for ChatGPT, so one missing tag on a URL is enough to make a whole channel look like it's collapsing.

If OpenAI, Google or Anthropic ever share proper analytics, we'd plug it straight into Obsero so customers get the full 360 view of AI search, from what the models say about them to the traffic they send. Until then, we'll keep using our own data to spot changes like this as they happen.

One last thing. Thank you for all the kind messages after Wednesday's news that Obsero has raised $1.2 million. If you missed it, you can read it here.

That is it for this Sunday, until next time.

Have a good one.

Andy