Step 1: fix your question set and keep it stable
Write 15 to 25 questions your customers actually ask, phrased the way they ask them. Mix three types: discovery questions ("best interior fit-out company in Dubai"), comparison questions ("is X better than Y for villa projects") and problem questions ("who can renovate a restaurant without closing it"). Add Arabic versions of the important ones, since AI tools answer differently in Arabic and your Arabic-speaking customers hear a different set of recommendations. Then freeze the list. The whole value of monthly tracking comes from comparing like with like, and every time you swap questions mid-stream you break the trend line and start the clock again. Revisit the set at most twice a year, and write down exactly what changed and when, so you can still explain a jump or dip six months later.
Step 2: run the questions the same way each month
Pick a fixed week of the month and run every question through ChatGPT, Google AI Overviews, Gemini and Perplexity. Use fresh chats with no history, because previous conversations color the answers, and stay signed out of anything that personalizes results where possible. Record the full response rather than just whether you appeared: which businesses were named, in what order, and what the answer actually said about each one. Screenshots help when you want to show a partner or your team exactly what AI says about your brand instead of a summary of it. Expect the full run to take an hour or two for a 20-question set once the routine settles into a rhythm.
Step 3: score the month in one spreadsheet row
Your Brand Mention Rate is the share of answers that name your business. Count each question-platform pair as one answer: 20 questions across four platforms gives 80 answers, and if you appear in 12 of them your rate is 15 percent. Track three more columns beside it: your average position when named, the competitors named most often, and any factual errors in how you were described. Wrong prices, dead links and outdated addresses inside AI answers are early warnings, and each one should get logged as a fix on your website or directory listings. Those numbers fit in one row per month, which is the entire reporting overhead, and a simple line chart of the rate is usually all a management meeting needs to see.
Step 4: read trends, not single months
AI answers move around naturally, so a two-point dip in one month means little, while three months of steady drift means something changed. Watch for step changes after your own fixes go live, after a competitor lands press coverage, and after the platforms update their models. When a competitor suddenly climbs, read the answers closely and note which new sources they cite. That source is usually why they got the lift, and it tells you where to go next. A professional AI visibility baseline applies this same discipline across markets and languages. Doing it monthly is what turns a one-time snapshot into a number your team can actually manage against.