An audit is not a feeling about how your brand is doing. It is a repeatable procedure that produces the same kind of number every time so you can compare quarters. Ninety minutes, once every three months, and you will know more about your public identity than most people know about theirs.
One rule before you start: do not fix anything while you are auditing. The urge to stop and correct a typo is strong and it will cost you the audit, because you will end up half-fixing three things and never finishing the diagnosis. Write it down and move on.
01Setup
- Open a private or incognito window. Sign out of everything.
- Create a folder named with today's date. Everything goes in it.
- Have last quarter's folder open beside you if this is not your first run.
- Set aside ninety uninterrupted minutes.
02Check 1: the bare name search
Source: Our own seven client records, read live from Google on 31 August 2026
Search your name exactly as somebody would type it. Screenshot the full first page, including the bottom. Count how many of the ten results you control or would be happy for a client to read.
Score 2 if eight or more, 1 if five to seven, 0 if fewer than five.
03Check 2: paired queries
Search your name plus your city, your name plus your field, and your name plus your company. Screenshot each. These reveal namesake problems and tell you whether your differentiators are actually attached to you.
Score 2 if you dominate all three, 1 if you win two, 0 otherwise.
04Check 3: name consistency
| Name | Entity ID | Stage reached |
|---|---|---|
| Harsh Pamnani | /g/11f0z6v3zr | Stage 2 — panel fires on the bare name |
| Rohit Bassi | /g/11gbh01wq4 | Stage 2 — panel fires on the bare name |
| Wanuri Kahiu | /m/09gps9j | Stage 2 — worked on, not built |
| Himani Kankaria | /g/11rggfbxt1 | Stage 1 — entity record live |
| Katerina Khramova | /g/11f0zrgtrz | Stage 1 — entity record live |
| Sagar Kathrotiya | /g/11ld4h_jsy | Stage 1 — entity record live |
| Ramesh Pattni | /g/11kpwvj02p | Stage 1 — entity record live |
Open every profile you hold and every third-party record you know of. Compare the name form, the photograph, the job title and the one-line description against your written standard.
- Same name form everywhere, character for character
- Same photograph, or at least the same person recognisably
- Same job title wording
- Same one-line description, or near enough that a reader would not notice a difference
- Every one linking back to your site
Score 2 if all consistent, 1 if minor drift, 0 if you found different name forms.
05Check 4: the entity home
Open your own site. Select all, copy, paste into a plain text editor. Read what appears.
Score 2 if your full name, role, organisation and city are all present as text and the about page is over four hundred words. Score 1 if some are missing. Score 0 if your name is not in the text at all, which happens more often than you would think on well-designed sites.
06Check 5: structured data
Source: Our own seven client records, read live from Google on 31 August 2026
View the page source of your homepage and about page and look for a Person block in JSON-LD. Check that the facts in it match the visible text, and that the sameAs array lists your live profiles and nothing else.
Score 2 if present, valid and accurate. 1 if present but incomplete or stale. 0 if absent.
07Check 6: the proof layer
List every piece of verifiable proof a stranger could find without asking you: published writing under your byline, talks with a public record, press mentions, books, research.
Score 2 if four or more genuine third-party records exist and are findable. 1 if one to three. 0 if none.
08Check 7: the AI answers
Ask ChatGPT, Claude and Perplexity the same three questions, in a fresh conversation each: who is [your name], what is [your name] known for, and who are the leading people working on [your field] in [your city].
Save every answer verbatim with the date and the model. Note specifically: is it accurate, is it confident, does it confuse you with someone else, and does the third question name you unprompted.
Score 2 if all three describe you accurately and at least one names you unprompted in the category question. 1 if descriptions are accurate but you are never named unprompted. 0 if any model gets you wrong or confuses you with a namesake.
09Check 8: technical health
- Your pages return a clean status directly, without a redirect in the way
- Each canonical matches its own live URL exactly, including the trailing slash
- Your sitemap lists only URLs that resolve directly
- Search Console shows no coverage errors on your important pages
- The site works properly on a phone as its own layout
Score 2 if all clean, 1 if minor issues, 0 if pages are redirecting or missing from the index.
10Check 9: the knowledge panel state
Note whether a panel appears on your bare name. If it does, read every line and log any errors, and check whether you have claimed it. If it does not, note which requirement is thinnest: a weak entity home, insufficient third-party coverage, name inconsistency, or simply not enough time yet.
Score 2 if a panel exists, is accurate and is claimed. 1 if it exists with issues. 0 if none.
11Scoring and what to do with it
| Total | Reading | Next move |
|---|---|---|
| 15 to 18 | Strong. Maintenance mode. | Protect consistency, keep earning records |
| 10 to 14 | Working. Gaps are specific. | Fix the two lowest checks only |
| 5 to 9 | Foundation incomplete | Stop all output, spend a quarter on checks 3, 4 and 5 |
| 0 to 4 | Effectively invisible | Start at the name standard and the site |
The audit discipline
- Same wording, same setup, same order, every quarter
- Signed out, always
- Save everything, dated
- Never fix while auditing
- Compare to last quarter before deciding anything
The single number matters less than which checks moved. A score that went from eleven to twelve while check seven fell from two to zero is not a good quarter, and only the itemised comparison shows you that.
What to remember
- Audit signed out in a private window, or you are grading a version nobody else sees.
- Nine checks, ninety minutes, scored out of eighteen, identical every quarter.
- Never fix while auditing. Write it down and move on, or you will finish neither job.
- A model stating something false about you is more urgent than a model not knowing you.
- The value is in which checks moved, not in the total.
Questions people actually ask
How do I audit my own personal brand?
Run nine checks in a signed-out private window: the bare name search, paired queries, name consistency across profiles, whether your site's text contains your key facts, your structured data, your proof layer, what the AI assistants say about you, technical health, and your knowledge panel state. Score each out of two and repeat identically every quarter.
How often should I audit my personal brand?
Quarterly. More often and you will be reacting to normal fluctuation in search results, which changes day to day for reasons nobody outside Google can explain. Less often and problems like name drift or a wrong AI answer run unchecked for too long.
What should I check first in a personal brand audit?
Whether your own website's plain text contains your full name, role, organisation and city. Select all on your homepage, copy it into a text editor and read what appears. This single check catches the most expensive and most invisible failure, and it takes thirty seconds.
How do I check what AI says about me?
Ask ChatGPT, Claude and Perplexity the same three questions in fresh conversations: who you are, what you are known for, and who leads your field in your city. Save every answer verbatim with the model name and date, and watch particularly for confident wrong answers, which are more urgent than being unknown.
What is a good personal brand audit score?
On the eighteen-point version here, fifteen or above means maintenance mode, ten to fourteen means specific fixable gaps, and below ten means the foundation is incomplete and output should pause until checks three, four and five are solid. The trend between quarters matters more than any single total.
