Al's visibility isn't a number until the end of the month. ChatGPT, Perplexity, Google AI Overviews answers to the same question every week -- changing models, retrofitting sources, creating new content, your brand could disappear from the answer. This is why monitoring is done on a weekly frequency and only on eight indicators: more people are not watching, less can be seen.
Why do you monitor the weekly frequency?
The answer of the AI engine is not static. The model fine-tunes every few weeks, and the weight of capturing the source is moving; a new comparative article from the competition may replace the place you were quoted. We've seen the same hint in our surveillance of our clients, and within two weeks the brand fell from "first recommended" to "none at all" and there was no algorithm announcement. The monthly report will only show you four weeks later, and the weekly report will show you the signal on the seventh day, and you'll get the content.
8 indicators tracked every week
Not as many indicators as possible. There's too many columns, no one fills in and nobody reads. The following 8 are indicators that we actually use, each of which can respond to a specific action. Ask with a set of fixed 15 to 25 hints that run on the same day every week to compare numbers.
- Brand reference rate: In your fixed hint, AI answers the proportion that refers directly to the brand name. This is the baseline line.
- Number of hints: How many of the tracking hints trigger brands. From 3/20 to 12/20, it's the breadth of visibility.
- Cross-engine over: The same problem, you showed up several times in ChatGPT, Perplexity, Gemini, Google AI Overviews. A single platform is good. It doesn't mean it's good.
- Source citation: AI answers the ratio of your domain to the source when attaching the source link. It was mentioned, not to be quoted.
- Quoted page: the physical URL where they actually came from. Knowing which pages work for you, you know which one to copy.
- Description Accuracy: How AI describes your product, price and location, and whether old or competitive information is given to you.
- Share voice: In the same set of hints, your references are in proportion to all brands and can be directly compared to competition.
- In the order: you're in the first place on the list. The first and fifth clicks and memories are very different.
Three-level interpretation: appearance, quality, competition
The first three points answer one question: AI can see you or not. Brand reference rates are the baseline, the hint covers the breadth of your visibility, and the cross-engine overlay avoids you being deceived by a single platform. A lot of brands show up in Perplexity, in Google AI Overviews, but the whole thing is blank -- the two have different origins and separate pursuits.
Three indicators in the middle deal with the "when seen" problem. It's not like it's being quoted. It's not necessarily the page you want to play. If you write down the quoted URL, you will find that what gives you visibility is not a home page or a product page, but a specific teaching or comparative article. The accuracy of the description is even more important: AI says your old price or competitive function three years ago is more lethal than not appearing.
Last two indicators put you back on the market. Sharing volume is the proportion of your references in the same set of questions, which are relative numbers, which compare directly with competition. The order of appearance reflects the level of trust that AI has in you -- it's ranked first and fifth on the list, and there's a big difference between the hits and the memories of the readers. These two indicators, taken together, will determine whether you are in the lead or are being pursued.
What does the weekly template look like?
Put the finger sign down as a weekly, the simpler the structure, the better. One page is enough for any colleague to see what happened this week in three minutes.
- Header: This week's date, monitored hints, covered AI platform.
- This week 's values for 8 indicators, as well as increases and decreases from the previous week, can be marked with + and - no fancy graphs.
- Three highlights: the most cited page this week, the most progressive hint, the most missing hint.
- A conclusion: an action this week, such as adding a piece, correcting a misdescription, chasing a competition.
- Appendices: At least three AI responses, original screenshots or full text, traceable later.
Three common mistakes that invalidated the weekly.
First, change the hint every week. As soon as the hint moves, the numbers lose their comparability, which is tantamount to restarting; fix a set and run at least one season before the examination. Second, just watch ChatGPT. It's high-market, but your potential clients may be more likely to use Perplexity for comparison, initial selection with Google AI Overviews. Third, only numbers don't remember action. If the weekly is not "so what to do this week", it is just a report that will be filed and will not change anything.
Starting with the first weekly.
We don't have to wait until the tools are in place. Pick out 15 of your favorite questions, ask first round this week about ChatGPT, Perplexity and Google AI Overviews, fill in the table with eight points, and you have the first baseline. If you want to know the full gap in the AI answer now and which pages should be better, you can expect 30 minutes of GEO diagnosis, and we'll use Brand Radar to measure these eight indicators once.
