AI & Automation

Why does Google AI Mode give different answers every time you check your brand? Google launched temporary threads on Nov 17, with no history and no personalization: SMBs can run a monthly 20-question brand visibility check with it before paying €75+ a month for GEO tracking tools

2026.11.18 · 81 views
Why does Google AI Mode give different answers every time you check your brand? Google launched temporary threads on Nov 17, with no history and no personalization: SMBs can run a monthly 20-question brand visibility check with it before paying €75+ a month for GEO tracking tools
“

On November 17, Google added temporary threads to AI Mode: conversations are not saved, answers ignore past preferences and do not shape future personalization, rolling out globally on desktop and mobile browsers in every AI Mode language. With AI Mode past 1 billion monthly users and personalization expanding, owners checking their own brand often see personalized answers. We cover the update, GEO tracking tools, its limits, and a subscription-free monthly brand visibility check. ”

Share:

Direct answer: Use AI Mode's temporary threads, launched by Google on Nov 17. They ignore your search history and preferences, so answers sit closer to a baseline. Run 20 brand questions every month, log whether you are mentioned and which URL is cited, and decide after three months whether a €75+/month GEO tool is worth it.

On November 17, Google announced temporary threads for AI Mode. Tap the icon in the top-right corner of AI Mode and the conversation is not saved to your account, answers do not use your past preferences, and nothing affects future personalization. Close it and it is gone. The feature is rolling out globally on desktop and mobile browsers in every AI Mode language, Taiwan included. 9to5Google notes the greeting switches to "Welcome, stranger" and the model picker still works.

Why does this matter for marketing? AI Mode is now mainstream. At I/O 2026 in May, Google said AI Mode passed 1 billion monthly users a year after launch, with queries more than doubling every quarter. Personal Intelligence expanded to nearly 200 countries and territories in 98 languages, optionally connecting Gmail and Photos. After Gemini 4 Hydrogen launched on November 12, answers became longer and broke questions into more sub-queries. The deeper the personalization, the more the same question differs from person to person. An owner checking "recommended system developer in Taichung" from their own account may see results shaped by their own browsing, so the conclusion is unreliable.

A whole category sells AI visibility monitoring. The Next Web reported that Berlin's Peec AI passed $10M in annual recurring revenue 16 months after launch, with plans from €75 a month covering 115 languages; rivals include Profound, OtterlyAI, Bluefish and Semrush. Analysts size the market at about $1B in 2025 and up to $17B by 2034. The category is growing, but every tool assumes it can simulate "a user asking," and personalization keeps loosening that assumption.

For Taiwan SMBs, the practical meaning is simple: you finally have a free, official way to see AI answers your own history has not touched, without opening an incognito window. Below: the update, what three readers should do, a tool comparison, the limits, and a subscription-free monthly check.

Event details: temporary threads remove personalization from the answer

The core change is that one AI Mode conversation is fully detached from your account.

  • No history: The thread does not appear in AI Mode history and cannot be reopened.
  • No personalization: Answers ignore past preferences and do not change future recommendations.
  • Access: The temporary-thread icon in AI Mode's top-right corner; no settings change needed.
  • Availability: Desktop and mobile browsers, global, all AI Mode languages.

Winners are users who want clean lookups and brands that want to know what a stranger sees. Monitoring reports that read visibility from signed-in, personalized accounts will now face an easy comparison.

What three readers should do now: write a 20-question brand set

Everyone should turn "how customers ask" into a fixed question set this week.

  • Brand owners and SMBs: Write 20 questions across service plus location, cost, comparison and recommendation, such as "booking system developer in Taipei" or "how much does a booking system cost." Run them in a temporary thread the first week of each month.
  • Marketers and SEOs: Log four fields per question: mentioned or not, cited URL, position, and competitors shown. Compare month over month.
  • Developers and agencies: List cited pages and check for clear price ranges, FAQs and structured data. Rewrite service pages that are never cited.

GEO tool comparison: tools win on volume, manual checks win on cleanliness

Paid tools cover more questions more often; manual temporary threads are free and closest to a non-personalized baseline.

OptionPriceStrengthBest for
Peec AIFrom €75/month115 languages, multi-platform trackingBrands with a marketer
ProfoundMostly enterprise plansLarge prompt sets, competitor analysisMid-to-large brands
OtterlyAILower entry plansSimple setupSmall teams
Semrush AI toolsRequires Semrush subscriptionFits existing SEO reportsTeams already on Semrush
Manual temporary threads plus spreadsheetFree, about 2 hours a monthOfficial, non-personalized baselineBudget-limited SMBs

What they won't tell you: a baseline is not the truth

The feature is useful, but two limits matter.

  • Real customers are personalized: Most people search signed in. A temporary thread shows what a stranger sees, not what your repeat customers see.
  • One run proves nothing: AI answers vary; the same question asked three times can reorder results. Watch at least three months of trend.
  • Location still matters: Temporary threads remove personal preferences, not location. "Recommended" queries can differ between Taipei and Taichung.

The no-subscription SMB alternative: a 2-hour monthly visibility check

Free tools are enough for a trackable routine.

  • Step 1: Build the 20-question set in Google Sheets with columns for question, date, mentioned, cited URL and competitors.
  • Step 2: On the first Monday of each month, ask each question twice in a temporary thread and record the majority result.
  • Step 3: Run the same set once in ChatGPT's temporary chat and in Bing to compare.
  • Step 4: Put never-cited service pages on next month's rewrite list: add prices, audience and FAQs.

ScriptWalker (a Taiwan Laravel/Flutter custom development studio) recommends three months of data, and evidence that visibility moves inquiries, before paying for a tool.

FAQ

What are temporary threads in AI Mode?

AI Mode conversations that are not saved or personalized and disappear when closed, rolling out globally from Nov 17.

How is it different from Incognito?

Neither keeps history, but temporary threads need no new window and keep AI Mode's model picker.

Is a temporary thread answer what customers see?

Not exactly. It is a non-personalized baseline; signed-in, personalized users may see different results.

Do SMBs need a GEO tracking tool?

Not yet. Run 20 brand questions monthly for three months first, then consider tools from €75 a month.

If my brand is missing from AI answers, what do I fix first?

Add price ranges, who it is for, and FAQs to service pages so AI has sentences it can quote.

My take: visibility scores will get less reliable; cited URLs won't

The mainstream view is that GEO tracking will become standard, like rank trackers once did. My view: as personalization grows, a score for "where your brand ranks in AI answers" will drift further from reality, because there is no single shared answer. By launching temporary threads, Google is effectively admitting that personalized and baseline answers are different things.

Two metrics stay stable: which of your URLs AI cites, and how many inquiries those pages bring. The first tells you which pages to fix; the second tells you whether fixing them pays. A rising visibility score with flat inquiries is a vanity metric.

For ScriptWalker, the opportunity is a small monthly service: a 20-question AI visibility check, a cited-page audit and service-page rewrites. Clients do not need a dashboard; they need to know which three pages to fix each month.

Sources

Share:
AI & Automation Back to Blog