1. Why this question
Everything we do begins with a name. Someone asks what Google and the AI models say about them, and before we can answer we have to be certain which person we are looking at. Get that wrong and every finding after it is about a stranger.
For a rare name this is trivial. For John Smith it is the entire problem. So the practical question is: what is the second word? Ask for a job title, an employer, a city, an industry? Every form on the internet asks for one of these. We wanted to know which one actually works, because ours was going to ask for something and we would rather it was the right thing.
2. Method
We took two people whose real online properties we already know, exhaustively, because we built or audited them: their own sites, their profiles, their register entries, their coverage. That gives a ground truth, which is what most search studies lack.
For each subject we ran five Google searches from the United Kingdom, logged out, on desktop: the bare name, then the name plus each of four modifier types. For each search we counted how many of the first ten organic results genuinely belong to that person.
Separately we searched three deliberately common names with no ground truth available, and recorded how concentrated the first page is and whether a knowledge panel fires. That measures the size of the problem rather than the fix.
All searches were run on 15 August 2026 through the DataForSEO live SERP API. Search results change, so treat every figure here as a snapshot with a date on it rather than a constant.
3. Results
Precision at 10 for each subject and modifier. Higher is better.
| Query type | Subject A | Subject B | Average | |
|---|---|---|---|---|
| Name plus industry | 80% | 40% | 60% | |
| Bare name, no modifier | 60% | 50% | 55% | |
| Name plus location | 60% | 40% | 50% | |
| Name plus job title | 60% | 30% | 45% | |
| Name plus company | 40% | 40% | 40% |
Subject A is a founder in communications. Subject B is an executive chairman in technology. Both are consenting parties whose properties we can verify. Neither is named here, because naming a client in a research paper is not something we do without a specific reason.
4. Three findings
Adding your job title makes it worse
Job title averaged 45 percent against 55 percent for the bare name. It is the single most commonly requested field on any "find yourself" form, and in our measurement it actively degraded the result. Titles are generic. "Founder" and "Executive Chairman" describe hundreds of thousands of people, so adding one broadens the field rather than narrowing it.
Adding your company is the worst option of the four
40 percent, fifteen points below simply searching the name. This is the most counter-intuitive result here, because your employer feels like the most identifying thing about you. What actually happens is that the search pivots towards the company: its website, its filings, its coverage, its other staff. You get a good page about the business and a worse page about the person.
Only industry beat the bare name, and not reliably
Industry averaged 60 percent, five points above the bare name, which sounds like a winner until you look at the split: 80 percent for one subject and 40 percent for the other. It worked brilliantly for the person whose industry term is strongly associated with them, and did nothing for the person whose is a widely used technical term. A modifier that helps one person and not the next is not a mechanism you can build a product on.
5. The common name problem, measured
Three deliberately common British names, searched with no modifier at all.
| Name | Distinct domains in top 10 | Knowledge panel |
|---|---|---|
| John Smith | 9 of 10 | Yes. A former Minister of State for the Privy Council Office |
| David Jones | 9 of 10 | None |
| Sarah Williams | 9 of 10 | Yes, described as another person's wife |
Nine distinct domains out of ten results means almost no concentration at all. The first page is not about a person, it is about a string of characters that many people happen to share.
The panels are the sharper lesson. Search John Smith and Google confidently shows you a politician. Search Sarah Williams and it shows you someone identified by who they are married to. If your name is Sarah Williams and you run a company, the machine-readable answer to "who is Sarah Williams" is currently somebody else's spouse, and no amount of adding your job title to a search box changes that.
6. So what does work
The thing all four modifiers have in common is that they are descriptive. They tell a search engine what kind of person you are, and it responds with people of that kind.
This is also the mechanism behind our other paper. What makes a source count in a machine readable record is not how well known it is but how unambiguously it points at one entity, which is why a public register outperforms an open wiki. What makes a trusted source sets out that evidence.
7. What we changed because of this
Research that does not change anything is decoration, so for the record:
- Our free audit does not ask for your job title. It asks for your name and one link that is definitely you, and the field explains why. We were going to ask for a title and a company, which this study says would have been the two worst choices available.
- The audit reports precision explicitly. Every report states how many of the first ten results are actually the subject, because that number turns out to be the honest headline for how resolved a person is.
- We treat a shared name as a finding, not a footnote. If someone else dominates your name, that is the first thing your report says.
The free audit is here if you want to see your own numbers.
8. Limitations, stated plainly
- Two subjects. Ground truth is expensive: it requires knowing every property a person genuinely owns. Two is enough to show a direction and not enough to fix a coefficient. Treat the ordering as the result, not the exact percentages.
- One country, one day, logged out. United Kingdom, desktop, 15 August 2026. Results differ by location and change over time.
- An anonymous search is not your search. A logged-out query from a data centre sees less than you see signed in, which is the same measurement problem we describe in our other paper. It is why we never report an absence as proven.
- Modifier quality varies. "Digital PR" and "Bittensor" are not equally specific terms, and that difference is visible in the industry row. A larger study would control for term frequency.
We publish the limitations because a study of six searches per person presented as settled science would be exactly the sort of thing this company exists to argue against.