Learn

What we have actually measured

Not a blog. Every page answers a question somebody typed into a search box, and every page carries something we measured, with the date and the method. Including the measurements that went against us.

Written by Sam Allcock, founder of Ascribed·10 November 2026

Pillar one

Google knowledge panels

Google knowledge panels, and what actually produces one

What a knowledge panel is, where Google gets the facts inside it, and what we have measured actually causes one to appear. With dated captures, including the ones that went against us.

How to get a Google knowledge panel

The actual sequence, in the order that works, with the free steps named as free and the one step where failing cannot be undone.

Can you claim your own knowledge panel?

What Official Representative status actually gets you, what it does not, and why nobody outside can tell whether a panel is claimed.

A knowledge panel for a person, not a company

Why a personal panel works differently from a company one, why a distinctive name beats a big one, and why a panel showing for your name may not be yours.

Eleven days from nothing to a knowledge panel

The fastest result we have recorded, exactly what was done, and the part of it that went wrong afterwards.

Why your knowledge panel shows the wrong photograph

Google picks the image when your record does not carry one, and the licence rules behind supplying your own are not what most people assume.

You cannot check a knowledge panel with a scraper

Measured twice on real captures. The panel header, the subtitle and the photographs are never drawn in an automated capture of a results page.

Pillar two

Your name in search

Why somebody else comes up when you search your own name

What the first page of results for your own name is actually made of, how much of it usually belongs to you, and what moves it. Measured across 21 real audits, including the two our own checker got wrong.

How much of your own first page actually belongs to you

We counted it across 21 real audits. 20 of 21 people owned under a fifth of their own first page, and the way we count it understates ownership rather than flattering it.

What to do when somebody else has your name

Namesakes are the most common thing on a first page after your own profiles, and the one people most want removed. Removal is not available. Here is what actually works.

How to rank for your own name

The steps in the order that works, with the free ones named as free. Most people have not done the first two, and they cost an evening.

When a result on your name is somebody else entirely

How to tell which results are yours, why a matching name proves nothing, and the two occasions our own checker attributed the wrong person.

Why Google shows the wrong photograph for your name

Google picks the picture when your record does not carry one. Why it chooses what it chooses, and why supplying your own is harder than it sounds.

When four sources describe you four different ways

Most first pages do not carry wrong information so much as inconsistent information. What that costs, measured on a real audit, and why it is the cheapest thing to fix.

Pillar three

What AI assistants say

What ChatGPT, Gemini, Perplexity and Claude say when somebody asks about you

We put the same three questions about one person to four assistants on the same day. All four named him. One leaned on twenty two sources and another leaned on one. Measured, with the method and the cost.

Where AI assistants get their information about people

The hosts actually cited, from real runs. Public registers, professional profiles, journalist databases, review sites and the subject's own pages, in roughly that order of frequency.

Why ChatGPT and Gemini give different answers about the same person

Twenty two sources against one, on the same person, the same question and the same day. What that gap means and why checking only one assistant tells you almost nothing.

When an assistant answers about a different company with your name

Fifty one sources, none of them the company's own, and twenty citations belonging to seven other firms sharing the first word of its name. The sharpest version of the shared name problem.

When an assistant cites a source that does not exist

Two measured cases: a citation pointing at a web address that has never existed, and a citation pointing at a real page that does not mention the subject at all.

Can you change what AI says about you?

There is no submission form and no control panel. What can be changed is what is available to be read, and most of the sources that actually get cited are editable.

How to check what AI assistants say about you

The free version takes twenty minutes and is worth doing. What to ask, what to record, and the three states a result can be in rather than two.

Pillar four

Your company and you

Your company and your own name are two different problems

Work on a founder's record does not fix a company's, and on one recurring case it makes it worse. What we have measured on real companies, and how to tell which problem you actually have.

When the Companies House result for your firm is a dissolved company

Measured on a real firm: the register result ranking for their own short name was dissolved three years earlier, and the company that actually trades was not on the page at all.

Two Google Business Profiles for one company

Found on a real firm: two profiles, one telephone number, all the reviews on one of them, and neither claimed. The fastest way to lose a review history.

When your old company follows you to the new one

The public register publishes the link between a director and every company they have run. Why that carries history across, and why building a stronger personal record makes it worse rather than better.

When your company name is made of ordinary words

We measured this on our own parent company. Nine of sixteen platforms reported carrying it and not one of them named it. What a generic name costs and what still works.

When the company number you give out is not yours

A real example: the number supplied on an audit form belonged to a dissolved building company in a different county. The register is the truth and almost nobody checks it.

The company profiles a search engine cannot find

A real firm was told it had no LinkedIn page and no YouTube channel. Both existed and the LinkedIn page ranked on page one for its own name. The fault was ours, and the lesson is general.

Start with your own name

Everything here came out of running this on real people. The free audit is the same check.

Run my free AI audit