Ninety Days to a Different Answer
A quarter is long enough to change what the engines say about a person and short enough to be held to it. This is how the ninety days are spent, and what is counted at the end.
Personal reputation & personal branding for the AI era · Founder, ExcBrand.AI
5 min read
In one paragraph
A ninety-day reputation plan reads the answer first, fixes the sources it is built from, adds owned pages for the questions nothing answers, then earns a few credible placements. Every target is a query with a reading taken weekly, and a flag goes up when a reading falls behind. Day ninety compares answers, not activity.
Reputation work has a habit of being open-ended. A retainer starts, activity accumulates, and a year later nobody can say what changed, because nobody wrote down what it was meant to change. We work in quarters for the opposite reason: ninety days is long enough to move an answer and short enough that a plan has to say, at the start, what it will have moved.
Days one to fourteen: read
Nothing is written in the first two weeks. Each engine is asked the questions the shortlist asks, in the words they would use, and the answers are kept verbatim, including the parts that are wrong. The paired searches are read the same way. Then every claim in every answer is traced to the source it most plausibly came from.
This is the step most programmes skip, and it is the one that decides the rest. A plan written before the reading is a plan for the problem someone assumed.
Days fifteen to forty-five: fix and build
Fixing comes before building because it is cheaper and it compounds. A wrong title on a profile the person controls, an outdated biography on a company page, a directory entry that merges them with someone else: each is a source an engine may be reading, and many can be corrected by the person who owns them.
Building fills the gaps the reading found: a page in the person’s own name, structured so a machine can read it, that answers the questions nothing else answers. Few pages, each doing one job.
Days thirty to ninety: earn
Earning starts once there is something correct for it to point at. A small number of placements, each chosen to answer a specific query in the set, on sites the shortlist actually reads and whose editors would publish the piece on its merits. Each piece carries the label it earned.
Targets and flags
Every target in the plan is written against a query, not an activity. Not “publish four articles” but “when the shortlist asks who leads this company, every engine we read names the person and their current role correctly”. A target like that can be checked every week, by anyone, against the reading.
A flag goes up when a reading falls behind the line the plan expected for that week. A flag is a question, not an alarm: is the source still wrong, was the page not indexed, did the engine change what it cites. Each flag gets a cause and an owner before the next reading.
A plan that cannot be wrong by week six cannot be right by week thirteen.
Day ninety: compare answers
The report at the end of the quarter sets the first reading beside the last, for every query in the set, in every engine read. What changed, what did not, and what the next quarter should target. It does not count impressions, posts or mentions, because none of those is what the shortlist reads.
The terms this piece uses
- Reading
- One dated record of what an engine or a search page returned for one query, kept verbatim.
- Target
- The answer a query should return by a set date, written so that anyone can check it against a reading.
- Flag
- A marker raised when a weekly reading falls behind its target, answered with a cause and an owner before the next reading.
What to take from this
- A quarter is long enough to change an answer and short enough to be held to a target.
- Read before writing: the first two weeks produce no content, only verbatim answers and their sources.
- Fix sources before building pages, and build pages before earning placements.
- Every target is a query with a checkable answer, not a count of activity.
- A flag is a question with an owner, raised when a weekly reading falls behind.
- Day ninety compares the first answer with the last.
Questions people ask
- How long does it take to change what AI engines say about a person?
- We plan in quarters. Ninety days is enough to read the answers, correct the sources, publish owned pages and earn a few placements, with weekly readings showing whether the answer is moving. The engines re-read on their own schedule, so no single date is guaranteed.
- What should a reputation plan measure?
- The answer itself: for each query the shortlist asks, what each engine says, read weekly against a written target. Not impressions, posts or mentions.
- What is the first step of an AI reputation plan?
- Reading. Ask each engine the shortlist’s questions, keep the answers verbatim, and trace each claim to its likely source before writing anything.
Written with the help of AI and reviewed by our team.
More in Method
Fix the Source, Not the Model
When an AI answer gets a fact about you wrong, the error almost always comes from a page the engine read. AI reputation management starts by finding that page, not by arguing with the model.
5 min read
When the Sources Disagree
A profile and its placements are read together. When they describe the same person differently, an engine has to choose between them, and it does not always choose the page you own.
4 min read
The Search That Pairs Your Name
The people who matter rarely type a name alone. They type it next to a company, a role or another name, and the search box finishes the thought for them. That pairing is the real unit of reputation work.
6 min read