Your visibility on AI tools drops sharply in February and you think you did something wrong — but it could be a documented seasonal cycle, more pronounced than the one in Google traffic. If you don't know about it, you risk changing strategy at the worst possible moment, exactly when all you need to do is wait. Learning to read the seasonal rhythm of AI visibility saves you from costly decisions based on a false alarm.
Across 4 brands I tracked for 12 months, visibility in AI answers shows seasonal cycles more pronounced than classic organic traffic. From January to March citations rose by 40%, in June-July they fell by 25%. Understanding the rhythm means understanding when to act, and above all when NOT to panic over a drop that is merely seasonal.
Let me explain with a case I experience firsthand. A seafood restaurant on the Grado lagoon, in the province of Gorizia, calls me in September, worried: “Roberto, in August ChatGPT and Perplexity were citing us for ‘where to eat seafood in the lagoon’, now they don’t cite us anymore. Did we do something wrong?”. My answer: no, you did everything right. The thing is that in September the query “seafood restaurants Grado lagoon” collapses, and with it the frequency with which AI engines regenerate answers that include their name. AI visibility is alive when the demand is alive.
In the previous articles in this series I showed you how to measure AI visibility in absolute terms — share of voice, citation accuracy, position. Now we add the axis of time. Without it, you risk changing strategy over a drop that was already on the calendar.
What “seasonality” means for an AI engine
Generative models don’t have a built-in calendar. But they respond to queries, and queries have seasonality.
In the field of research on user search behavior, it has been documented for years — Google Trends has existed since 2006 — that query volumes oscillate in yearly, weekly and intra-week cycles depending on the sector. From this it follows, by direct deduction, that AI answers also follow a cycle: if no one asks “where to eat seafood in Grado” in February, ChatGPT doesn’t regenerate that answer, and you don’t get “seen” even if you’re perfectly optimized.
Translated into practice: your AI visibility is the intersection of two curves — your authority as perceived by the model (static in the short term) and the query volume of your semantic field (seasonal). If you ignore the second, you misread the first.
Why AI cycles are more pronounced than search cycles
On Google, the drop in query volume in February translates into a drop in traffic, but SERP positions stay stable. On AI, however, something different happens: fewer queries → fewer citation opportunities → fewer relevance signals for that semantic area → a further drop in citations even for the few queries that remain.
It’s an effect of seasonal amplification that I’ve seen across all 4 tracked brands. The search drop was 15% in June-July, the AI citation drop was 25%. The autumn rebound was symmetrical: search +20%, AI +35%.
The reason, deduced from what we know about how models weigh the freshness and frequency of signals, is that AI not only “sees the query less”, but “sees the topic less” and therefore surfaces the names from that semantic area less often even in side conversations.
Overhauling the site or the tone every 3 months resets the signals and makes you lose ground exactly when the seasonal cycle was about to bring you back up.
The test you can run in 20 minutes
You need Google Trends and a spreadsheet. Nothing esoteric.
Open Google Trends, enter your sector’s main query — for the Grado restaurant it would be “seafood restaurants Grado” or “where to eat in the lagoon” — and select a 5-year range, Italy region. You get your sector’s seasonal curve. Note down:
- The three peak months (for seafood in the lagoon: June-August)
- The three trough months (January-February, November)
- The amplitude between peak and trough (if >50 points out of 100, you’re in a strongly seasonal sector)
Then, every week, do this: open ChatGPT, Perplexity and Gemini and ask the 5 main queries in your sector. Count how many times you’re cited. Note it down on an Excel sheet with the date.
Binary decision threshold: if your weekly drop in citations is in line with the seasonal Google Trends drop of your semantic field (within ±10 percentage points), it’s not a strategic problem, it’s the calendar. If instead your drop is 30% while Google Trends shows -10%, that’s when you really have a signal to investigate.
Plan your content 2-3 months ahead of the peak.
The test I ran myself
Over the 12 months that just ended, I tracked 4 brands across different sectors and territories: the seafood restaurant on the Grado lagoon, a Tuscan DOCG winery in the province of Siena, an accountancy firm in Bari, an artisanal pastry shop in Naples. One main query each, three engines (ChatGPT, Perplexity, Gemini), a weekly check.
The seasonal pattern emerged clearly across all 4, with different amplitudes:
- The Grado restaurant: peak-to-trough amplitude of 65% (strong tourism sector)
- The Siena winery: amplitude 45%, with a second peak in November (harvest storytelling + Christmas gifts)
- The Bari accountancy firm: amplitude 50%, peaks in May (income tax returns) and November (advance payments)
- The Naples pastry shop: amplitude 40%, peak in December (artisanal Christmas sweets)
Limitation of the test: a sample of 4 brands, one query each, 12 months. It’s not a study, it’s a longitudinal observation that confirmed a hunch. Real analysis requires professional AI tracking tools with historical data and sector benchmarks. But even with this minimal setup I was able to tell the 4 clients “you don’t have a problem, you’re in a trough” or “you have a real problem, let’s act”.
The most common and widespread mistakes
Mistaking the July drop for an SEO disaster. It happens every year with tourism clients: in July the owner is busy in a packed venue and doesn’t look at the tools, in August they get back to the desk, see the citation drop from the mid-August holiday week (when they themselves were closed) and panic. It was seasonality, worsened by their own closure.
Launching content at the wrong time. Publishing a “guide to Tuscan wines for Christmas” article in December is too late — the model has to see it, index it, and the signals have to accumulate. The publication that’s useful for the Christmas season needs to go out in September-October, exactly as with classic SEO but with a wider window, because generalist models update less often than Google’s indexes.
Comparing yourself only to yourself. If your citations drop by 20% in February, but your 3 direct competitors drop by 30%, you’ve gained share. Without comparison to the sector, you misread even a positive result. Always compare against the 3-5 competitors that AI cites in your field.
Changing strategy every three months. AI engines reward the consistency of signals over time (I went deeper on this when talking about implicit reference weight). Overhauling the site or the tone every 3 months resets the signals and makes you lose ground exactly when the seasonal cycle was about to bring you back up.
What can we concretely do?
Three actions, in order:
- Build your sector’s seasonality curve with Google Trends over 5 years. It’s your reading benchmark for the next 12 months.
- Weekly AI tracking with an Excel sheet for at least 6 months before drawing conclusions about non-seasonal trends. With 4 weeks you see nothing, with 12 you start to read it, with 26 you can tell noise from signal.
- Plan your content 2-3 months ahead of the peak. The Grado restaurant publishes its “summer 2026 menu in the lagoon” pages in March, not in June. The Siena winery writes its Christmas gift pages in September.
For the entity coverage of your seasonal content, what I told you about entity recognition in text will come in handy: if AI doesn’t recognize “Grado”, “Marano lagoon”, “Chianti Classico” as entities connected to your brand, even the seasonal peak brings you fewer citations than it could.
Seasonality and visibility in AI answers
Maturity in measuring visibility in AI answers comes from understanding that it isn’t a flat metric. It oscillates with demand, with the calendar, with sector events. Without the 12-month axis of time in front of you, every decision is a reaction to noise.
In the upcoming articles in this series I take you forward on two connected fronts: the correlation between AI visibility and classic SEO visibility, to understand how much the two curves move together, and AI visibility forecasting, to use seasonality not just to read the past but to predict the next 90 days.
