Brand Authority

Brand authority assesses how present a brand is beyond its own website - on the platforms AI systems draw their answers from. It contributes 20% to the GEO score.

Mentions without a link influence visibility in AI systems more strongly than classic backlinks. What matters is not the volume but where the brand is mentioned: one Reddit thread about experiences with your product counts for more, to a language model, than ten directory listings.

Which platforms are checked

The dimension is not a single value but a weighted mix of individual platform checks. The weights are ratios, not percentages - they only describe how strongly a platform counts compared to the others.

Platform	Weight	Source of the value	Default
YouTube	18	measured	on
Press	16	measured	on
Wikipedia	15	measured	on
Reddit	14	measured	on
LinkedIn	12	self-declared	on
Wikidata	10	measured	on
Reviews	10	measured	on
Podcasts	8	measured	on
Quora	5	self-declared	on
GitHub	12	measured	off
Stack Overflow	8	measured	off
Hacker News	6	measured	off

GitHub, Stack Overflow and Hacker News are off by default. For most brands a missing GitHub organisation says nothing. For technical products, enable the three in the scan profile.

How a platform value is built

Every platform is scored from 0 to 100 on its own. The value is made up of partial points that always follow the same four questions: does the presence exist at all? How much happens there? How widely is it spread? How recent is it?

Partial points are awarded in steps, not linearly. The difference between 200 and 400 subscribers says nothing about visibility in AI systems, the difference between 200 and 200,000 says a lot.

Platform	Partial points
YouTube	channel for the brand 30 · upload within the last 90 days 20 · subscribers up to 25 · videos by other channels up to 25
Press	article within the last 90 days 40 · number of reporting outlets up to 30 · volume over six months up to 30
Wikipedia	article with a matching title 40 · length and maintenance up to 20 · mentions in other articles up to 20 · domain cited as a source 20
Reddit	mentions present 30 · volume up to 30 · number of different subreddits up to 20 · dedicated subreddit 20
LinkedIn	linked company page 60 · additionally linked personal profile 40
Wikidata	entity for the brand 40 · official website matches the domain 25 · properties up to 20 · language versions up to 15
Reviews	Trustpilot profile 40 · number of reviews up to 35 · average rating up to 25
Podcasts	own show 40 · episodes of other shows up to 40 · most recent mention within six months 20
Quora	linked profile 100
GitHub	organisation for the brand 40 · repositories and activity 30 · stars up to 30
Stack Overflow	dedicated tag 40 · volume of questions up to 40 · share of answered questions up to 20
Hacker News	volume of stories up to 50 · sum of points up to 30 · most recent story within a year 20

A few decisions behind this deserve an explanation:

• Diversity counts more than volume. Several subreddits or several newsrooms talking about you weigh more than many posts in the same place - it shows reach beyond your own niche.
• Third-party statements count more than self-presentation. Mentions in other people's podcast episodes count as much as running your own show. Your domain being cited as a source in Wikipedia is the single most telling signal of all, because it cannot be arranged by yourself.
• For press, recency counts most. Press is a perishable signal: the largest single item is an article from the last 90 days, not the total volume.
• A mismatched hit is not the same as no hit. If the Wikipedia search finds an article whose title does not match the brand, it earns a third of the article points instead of none. Conversely, if the official website of a Wikidata entity points to a different domain, the entity belongs to a similarly named organisation and does not count as a hit.
• Review volume counts more than the rating. A few enthusiastic voices carry less than many satisfied ones: a broad review base is the more reliable signal.

How the overall value is derived

1. Weighted average across platforms. The individual values and the weights from the table above produce a value from 0 to 100.
2. Self-declarations count half. LinkedIn and Quora cannot be queried; their value rests solely on the profiles your website links itself. A linked profile proves that it exists, not that anything happens there - so it enters at half weight.
3. Platforms that cannot be assessed drop out. Deselected, unreachable or without meaning for your brand: such platforms do not count as 0, they drop out of the calculation. Otherwise every newly added platform would depress the values of all existing projects, and a run with four platforms would not be comparable to a run with twelve.
4. The AI estimate is added. With AI analysis enabled, the measured part is combined with an assessment of brand visibility. Its share is 40% and drops to 20% as soon as at least four platforms have actually been measured. The more real measurements there are, the less should be estimated. Without AI analysis, only the measured part counts.

Where a value comes from

Every partial result states in the report what it is based on:

• Measured - queried through an official interface or public page of the platform.
• Self-declared - derived only from a linked profile, for example from the Schema.org field sameAs. Counts at half weight.
• AI estimate - a reasoned assumption without a measurement.

The difference between 0 points and not assessed matters. A 0 is a finding: something was searched for and not found. "Not assessed" means the question stays open - for example when the Reddit search was rate-limited. The report shows this in amber instead of red, and the value drops out of the calculation instead of counting against the brand.

What the report shows

The Brand presence card shows, for each platform, its logo, the value from 0 to 100, one sentence about the result and below it the metrics behind it - subscribers, number of threads, most recent article and so on. A badge only appears when the value is not a measurement of our own.

From the second run of the same scan profile onwards, every platform value and the dimension total state how they developed compared with the previous run. Above the list the numbers are summarised, for example "3 improved, 1 declined, 8 unchanged". A platform you have just enabled is marked "new" rather than as a rise from 0, and one that was assessed in the previous run but could not be determined this time keeps its earlier value on display. How the comparison works and which previous run it picks is described under GEO Audit. On the project tab the history chart can be switched to a single platform.

Next to it, the Profiles linked from the website card lists the platform profiles your website declares itself: from the Schema.org markup and from the links of the analysed pages. It therefore shows what the brand claims for itself, regardless of what is measurable there.

Settings in the scan profile

• Platform selection - every platform can be switched on or off individually. Deselected platforms are not queried and do not lower the value.
• Brand name - the basis of every search. Without it, the project name is used. A precise name prevents noise for generic terms.
• Founder name - optional. It adds a search for a Wikipedia article about that person. The hit does not feed into brand authority but into the assessment of experience and authority under E-E-A-T.
• Disable the brand search entirely - the dimension then stays not assessed and the GEO score is reweighted from the remaining criteria.

If the brand search is active but no platform at all is selected, the profile points this out. The run stays valid, but the dimension remains open.

For YouTube, an access key held by the platform operator is required to measure subscribers and third-party mentions. Without it, YouTube can reach at most 50 out of 100 points; the report notes this below the platform.

What does a low score mean?

A low score does not mean the brand is unknown. It shows that it could not be found clearly through the public signals that were checked. For local, new or highly specialized organizations, missing Wikipedia articles, podcast mentions or press coverage are normal.

How to improve the score

1. Use the brand name consistently on the website and enter it in the scan profile in exactly the same way.
2. Link verified official profiles through sameAs in the Schema.org markup - not just as an icon in the footer. Only then is the association explicit, for this audit as well as for AI systems.
3. Work towards third-party mentions: trade press, guest articles, podcast appearances, answers in your industry's communities.
4. Keep an existing presence current. A channel without an upload for a year and a press kit from two years ago are findings of their own in the action plan.
5. Maintain an existing Wikidata entity, in particular the property P856 (official website).
6. Ask satisfied customers for reviews. Volume carries more than the last tenth of a rating point.

Limits

The audit does not create entries and does not publish anything. Fabricated or purely promotional Wikipedia and Wikidata entries should never be pushed - they get removed there and harm the brand more than a missing entry.

Platforms that prohibit automated access are not queried. This applies to LinkedIn and Quora, which only enter as self-declarations, as well as G2 and Capterra, which only feed in through the AI estimate. For reviews, only the public Trustpilot profile is read.