Discussion
  • AI Search

Personal Intelligence lets AI Mode draw on a user’s Gmail and Calendar. How much does that personalisation undermine logged-out prompt tracking?

Published 1 reply

Discussion starter: an AI search question, posted with an answer from Niraj Raut to open the thread. If you have dealt with this on a site, reply with what you saw, especially where it differs.

Disclosure: Niraj Raut, who posted this answer, runs the SEO consultancy linked at the end of it.

1 reply

  • admin
    Admin
    Original poster

    It widens the gap between what a tracking tool sees and what users see, but it does not make logged-out tracking useless. Prompt tracking was always a sample from a variable distribution, not a record of what every user was shown. Personal Intelligence, which lets users who opt in give AI Mode access to context from services such as Gmail, Photos and Calendar, adds a new source of variation for those users. How much it matters depends on the category and on how many of your customers opt in, and Google has not published that number. Treat logged-out tracking as a baseline for non-personalised visibility, and rely more on outcome data to see what personalised users actually do.

    Where personal context is likely to change answers most

    Personal context is most useful to an assistant when the right answer depends on the person’s plans, history or preferences. That points to certain categories:

    Category Likely effect of personal context
    Travel and hotels High: booking confirmations, loyalty emails, calendar dates
    Restaurants and local services High: past reservations, location history, scheduled events
    Retail and shopping Moderate to high: past orders, brand emails, receipts
    Subscriptions and software Moderate: existing accounts and renewal emails
    B2B research Lower for most prompts, unless work accounts are connected
    General information Low

    These are reasonable expectations from how the feature is described, not measured effects. Google describes Personal Intelligence as opt-in, and its I/O 2026 announcements said it had become broadly available, but it has not shared adoption figures.

    What personalisation does to your tracking numbers

    Think of logged-out tracking as measuring whether you are in the candidate set when the assistant knows nothing about the user. For personalised users, the assistant may reorder or replace that set: a user whose inbox is full of a particular hotel chain’s emails may see that chain recommended ahead of others. So tracked visibility can overstate your presence for users loyal to competitors and understate it for users who already know you. On average across all users the errors may partly cancel, but you cannot assume they do.

    What still works

    • Logged-out tracking as a baseline. It still tells you whether you are eligible and competitive before personal context applies, and trends in it are still informative.
    • Search Console’s generative AI report. It counts impressions in AI Overviews and AI Mode for real users, personalised or not, although without clicks or queries.
    • Outcome data. Clicks from Google, branded search demand and conversions reflect what real users did after seeing personalised or non-personalised answers.
    • Customer research. Asking customers how they found you, including an AI assistant option, captures influence that neither tracking nor analytics shows.

    What not to do

    It is tempting to build synthetic logged-in personas with seeded inboxes to simulate personalised answers. I would be cautious. It is brittle, hard to make representative, may conflict with platform terms, and the results depend entirely on how you designed the personas. A small number of carefully designed personas can be a useful qualitative check on whether personal context changes recommendations in your category, but it is not a measurement system.

    The strategic implication: relationships become visibility

    If assistants draw on a user’s own emails, bookings and history, then the brands a person already has a relationship with gain an advantage in that person’s answers. That shifts some AI visibility work towards areas SEO does not usually own: loyalty programmes, transactional emails, account management and customer retention. A hotel group with a large, engaged loyalty base may be recommended to those members more often than tracking shows. A new entrant cannot use that route and has to rely on the non-personalised candidate set, which is what logged-out tracking measures.

    Hypothetical example:

    A regional hotel chain tracks 80 travel prompts logged out and appears in about a third of answers for its cities. Its Search Console generative AI report shows AI Mode impressions rising faster than tracked visibility, and bookings from Google clicks among existing loyalty members are growing. A plausible explanation is that personalised answers are favouring the chain for members whose inboxes contain its booking emails. The chain keeps logged-out tracking as its baseline for non-members, and adds a loyalty-member segment to its AI reporting built from bookings and branded search rather than prompts.

    Designing a prompt set that survives personalisation

    Some prompts are more exposed to personal context than others, and a tracking set can be designed with that in mind. Prompts that are explicitly generic (“best family hotels in Lisbon”) are the ones where a non-personalised baseline is most meaningful. Prompts that invite personal context (“find me somewhere to stay for my trip next month”, “what should I book for our anniversary”) are the ones most likely to diverge for opted-in users, and logged-out tracking of them measures little. I would tag each prompt in the set by how much it depends on the user’s circumstances, report the two groups separately, and put more weight on the generic group when judging whether the brand’s baseline visibility is moving. If the generic group holds steady while outcome metrics shift, personal context is a likely part of the explanation.

    How to report it

    Present logged-out visibility with a clear label: “non-personalised visibility across a fixed prompt set”. Alongside it, report AI impressions from Search Console and outcome metrics, and note that personalised users may see different answers. If leadership asks how much personalisation affects the numbers, the honest answer is that nobody outside Google can measure it directly yet, that it is likely to matter more in travel, local and retail than in B2B, and that the outcome data captures its effects even when tracking cannot.

    Keep logged-out tracking, label it honestly, lean more on outcome data, and recognise that in personal-context categories, customer relationships are becoming part of AI visibility.

    Need help with this on your own site?

    Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.

    See SEO services

Add a reply

Replies are open to approved members. Log in or apply to join.