 |
| System Overview (produced using ChatGPT) |
There's a moment many of us have had: you're looking at a competitor's website, then back at your own, and you just know something's different — but you can't quite put your finger on what. Traditionally, getting a rigorous answer meant hiring a consultant, running expensive user research, or spending hours doing it manually. What if you could get sharp, structured, comparative analysis in under ten minutes — without writing a single line of code?
That's exactly what this project set out to prove.
It started with a specific problem: comparing course websites to understand how they stacked up against a competitor. The goal wasn't just a surface-level look — it was to understand how real people, with different needs and backgrounds, would actually experience each site. The solution turned out to be a structured ChatGPT workflow built entirely in standard chat, using nothing more than a sequence of carefully designed prompts.
The core idea is simple but powerful: instead of asking ChatGPT one big question and hoping for the best, you break the task into stages. Each stage builds context before the next one begins. By the time the actual analysis runs, ChatGPT isn't working in the dark — it has a detailed picture of both websites and a clear human framework to evaluate them against. The result feels less like a generic AI summary and more like a considered brief from someone who actually did the reading.
Here's how it works.
Step 1 — Feed in the websites
The first subprompt instructs ChatGPT to ask for two websites, one at a time:
"Ask the user to enter two websites to compare. Label them as website1 and website2 respectively. Ask each one as a separate prompt."
Entering each site separately is deliberate. It gives ChatGPT a moment to analyse each one individually before any comparison begins — and it does. After each URL is entered, it produces a quick overview of key characteristics and early signals about the site's purpose, tone, and structure. Think of it as ChatGPT doing its homework before the debate starts.
Step 2 — Define your personas
This is where the workflow gets interesting. Rather than comparing websites in the abstract, the approach anchors the analysis in real human perspectives. Three personas are entered one at a time:
"Then ask three new prompts for new personas to be entered by the user. These will be labelled as persona1, persona2 and persona3 — a new prompt per persona."
The personas used in testing were deliberately varied: a time-pressed, university-educated man in his forties; a semi-retired woman with a doctoral background who leans towards world news; and a recently graduated engineer in his early twenties who lives on his phone. After each persona is entered, ChatGPT expands it — making reasonable assumptions about behaviour, expectations, and priorities. In testing, these inferences were consistently sensible and added useful texture to what could otherwise be quite flat demographic descriptions.
This step is worth pausing on, because it's the secret ingredient. Personas transform the analysis from "which site is better?" to "better for whom?" — which is a much more useful question to answer.
Step 3 — Run the analysis
With two websites and three personas loaded into context, the final subprompt does the heavy lifting:
"Please compare and contrast the websites against the personas. For each persona give a score out of 100 for the following: Overall score and Usability. Also for each persona add a summary. While analysing it take a pessimistic view and suggest improvements. Critically review the marketing and offer, and compare against each other. Present these in a graphical way to aid understanding. The audience to view the results of the analysis is the web team for the two sites."
The output is genuinely impressive. ChatGPT produced an executive summary for each site covering strengths, weaknesses and risks, followed by scored comparisons per persona. It then offered a strategic comparison across dimensions like trust, speed, content depth and engagement — ending with one-line recommendations per site. All without a single spreadsheet, survey or agency brief.
One instruction in that final prompt is worth highlighting: "take a pessimistic view." This small addition makes a meaningful difference. Left to its own devices, ChatGPT tends towards balance and diplomacy. Nudging it towards scepticism pushes the output past polite generalities and into the kind of direct, critical feedback that's actually useful for a web team trying to improve.
What worked well
The staged approach is what makes this work. Each subprompt doesn't just collect information — it primes ChatGPT to think in a particular way before the next input arrives. By the time the comparison runs, the model has a rich, structured mental model of both sites and all three users. That's fundamentally different from dumping everything into a single prompt and hoping for coherence.
The persona framework also proved its value. It gave stakeholders a way into the results that felt human and relatable, rather than abstract. A web team looking at scores for a 22-year-old engineering graduate will instinctively know what to prioritise in a way that a generic usability score simply doesn't communicate.
What's next
The workflow held up well, but there's clear room to evolve. The visualisations produced were functional but basic — future iterations should push for richer, more interactive outputs that make the data easier to present to senior stakeholders. More ambitiously, the analysis could be tailored so each persona receives a version of the report written for them — not just used as a lens within a single document. Imagine handing a one-page summary to a time-poor marketing director versus a detailed breakdown to a UX designer; the underlying data is the same, the framing entirely different.
There's also an argument for making the workflow more dynamic. Rather than moving linearly through the stages, a more sophisticated version might pause after the initial website analysis to ask clarifying questions, or allow personas to be weighted differently depending on the strategic priority of each audience segment.
Areas to Improve
- Better visualisations — move beyond basic outputs to richer, more interactive displays suited to senior stakeholders
- Persona-tailored reports — deliver each persona a version of the analysis written for them, not just referenced within a single document
- A more dynamic workflow — add clarifying questions mid-process and allow personas to be weighted by strategic priority
But as a starting point, this is a genuinely practical, no-code approach to competitive website analysis that any intermediate AI user can pick up today. The prompts are reusable, the structure adapts to almost any industry — from e-commerce to healthcare to financial services — and the whole thing runs in a standard ChatGPT session with no plugins, no integrations, and no specialist knowledge required.
Sometimes the most powerful tools are the ones hiding in plain sight. Have a go yourself and improve this. Have to see improvements via the comments.
Comments
Post a Comment