Research · Gibraltar AI Pulse
Gibraltar AI Pulse: do AI assistants agree?
Measured to · 1,107 answers to 140 questions
32%
Ask two AI assistants the same question about Gibraltar, and only about a third of the businesses they name are the same.
Averaged over 1,485 comparisons between ChatGPT, Claude, Gemini and Perplexity, each after searching the web. Being recommended by one says little about the others.
32%
of businesses named are shared between two assistants
13–33%
shared between an assistant's own knowledge and its web search
14%
of the websites cited are shared between two assistants
Which assistants agree most?
How much of their list two assistants share, for the same question, both after searching the web.
- Gemini and Google AI Overviews44%
- Claude and Gemini42%
- Gemini and Perplexity40%
- Claude and Google AI Overviews38%
- Google AI Overviews and Perplexity38%
- Claude and Perplexity37%
- Google AI Mode and Google AI Overviews36%
- Gemini and Google AI Mode36%
- Google AI Mode and Perplexity34%
- Claude and Google AI Mode28%
- ChatGPT and Gemini25%
- ChatGPT and Google AI Overviews23%
- ChatGPT and Perplexity21%
- ChatGPT and Google AI Mode21%
- ChatGPT and Claude21%
Does an assistant agree with itself?
ChatGPT, Claude and Gemini each answered every question twice: once from what the model already knows, and once after searching the web. This is how much the two answers share.
- Claude13%
- ChatGPT24%
- Gemini33%
The lower the figure, the more a business's place depends on whether the assistant happened to search. See why AI assistants give different answers.
Do they trust the same websites?
How many of the websites two assistants cited for the same question were the same.
- Claude and Perplexity22%
- Gemini and Google AI Overviews22%
- Gemini and Perplexity21%
- Google AI Mode and Google AI Overviews21%
- Claude and Gemini20%
- Google AI Overviews and Perplexity15%
- Claude and Google AI Overviews15%
- Gemini and Google AI Mode13%
- ChatGPT and Gemini12%
- Claude and Google AI Mode9%
- ChatGPT and Perplexity9%
- Google AI Mode and Perplexity9%
- ChatGPT and Google AI Overviews8%
- ChatGPT and Claude8%
- ChatGPT and Google AI Mode5%
Which sectors are most contested?
Agreement between assistants, by sector, least agreed first. Sectors with fewer than 20 comparisons are left out.
- Restaurants & Hospitality19%
- Construction & Trades19%
- E-commerce19%
- Cruise & Port23%
- Real Estate & Property23%
- Retail25%
- Accountancy & Tax Advisory26%
- Automotive27%
- Fund Administration29%
- Relocation & Residency29%
- Web Design & Digital Marketing30%
- Fintech & Payments30%
- Gaming & Betting32%
- Insurance33%
- Duty-Free Retail36%
- Health & Wellness37%
- Weddings37%
- Legal37%
- Education & Training47%
- Banking48%
- Yacht & Marina Services50%
- Hotels & Accommodation50%
- Yacht & Marina80%
How much changed since the last measurement?
Since August 2026, 53% of the businesses named in answer to the same question, by the same assistant, changed.
- ChatGPT, own knowledge75%
- ChatGPT, web search71%
- Claude, own knowledge57%
- Perplexity, web search50%
- Claude, web search49%
- Gemini, own knowledge41%
- Gemini, web search36%
What does this mean for a Gibraltar business?
- Checking one assistant tells you little about the rest. Measure each one.
- Web search changes the answer. Being findable on the web and being known to the model are two different jobs.
- Assistants trust different websites, so being listed in more of the right places widens your chances.
Common questions
What does agreement mean here?
For each question, we take the Gibraltar businesses two assistants named and divide the number they have in common by the number named by either. Two answers naming exactly the same businesses agree 100%; two with none in common agree 0%. We average that over every question both assistants answered.
Which businesses are compared?
Only businesses we track, because they are matched by their name and confirmed alternative names. Raw names are not compared: one assistant's "Sunborn Yacht Hotel" and another's "Sunborn Gibraltar" are the same place, and counting them as two would overstate disagreement.
How often is the Pulse updated?
After every measurement. We ask every published question once a month, so the Pulse moves monthly, and every figure says when it was measured. Month-on-month change appears once there are two measurements to compare.
Does low agreement mean the assistants are wrong?
No. Several different lists of good businesses can all be reasonable answers. It means that being recommended by one assistant says little about the others, so a business has to be measured in each of them.