LoudFace
Study · 17 November 2026 · Version 1.0

Google showed a knowledge panel for 21.9% of 899 software companies searched by name

How often Google shows a knowledge panel for a software company's name, and what goes with one: 899 companies from Wikidata (2026)

Sample
900 software companies drawn at random from 3,933 in Wikidata, 899 of them searched by name on Google (United States, desktop) on 17 November 2026, with 760 home pages read
Engines
Google Search, United States, desktop (via DataForSEO)

Data windowSearched and crawled 17 November 2026

21.9%

of 899 software companies got a Google knowledge panel for their name (95% interval 19.3 to 24.7)

Figure 1Share of software companies that got a knowledge panel for their name
Share of software companies that got a knowledge panel for their name
ValueNote
All 899 companies21.9%19.3 to 24.7
With an English Wikipedia article (252)49.2%43.1 to 55.3
Without one (647)11.3%9.1 to 14.0

Source: LoudFace study of 899 software companies in Wikidata, Google results via DataForSEO, 17 November 202617 Nov 2026

Key numbers

49.2% vs 11.3%

got a panel: companies with an English Wikipedia article, and without one

252 and 647 companies; every company has a Wikidata item
46.8% vs 2.6%

got a Knowledge Graph panel: with an English Wikipedia article, and without

Found after review: Business Profile panels counted apart; 252 and 647 companies
97.0%

of name searches showed an AI Overview

it cited the company's own site in 59.7% of them
10.0% vs 39.7%

got a panel: Wikidata records created since 28 August 2024, and older records

539 and 360 companies

Key takeaways

  • Google showed a knowledge panel for 21.9% of 899 software companies searched by name (95% interval 19.3 to 24.7); under a looser and a stricter definition, 24.8% and 20.5%.
  • Companies with an English Wikipedia article got a panel 49.2% of the time against 11.3% without one, a gap of 37.9 points that holds among unlisted companies (45.7% against 11.1%).
  • Found after review, by kind of panel: 46.8% of companies with an English Wikipedia article got a Knowledge Graph panel against 2.6% without; without an article, most panels were Business Profiles (8.7%).
  • Organization markup with sameAs did not go with more panels: found after review, within older records 43.1% with it against 40.9% without.
  • An AI Overview appeared on 97.0% of the name searches and cited the company's own site in 59.7% of them; 74.6% when a panel for the company was on the page.

When we typed a software company's name into Google, one of three things showed on the right of the page: a knowledge panel about that company, a panel about something else with a similar name, or nothing. We drew 900 software companies at random from Wikidata's open list and searched each name once from the United States on 17 November 2026; 899 searches returned. Kalicube is often quoted as finding that 45% of the brands its platform tracked had a knowledge panel in November 2020, up from 43% in May 2020; that frame is the brands its own users chose to track, and we could not find the figure on its live page or in an archived copy. Rothschild, Lurie and Mustafaraj (2019) noted that many news sources had no panel when searched, without a share. We found no measurement of this share for software companies, or for companies in general, on an open sample anyone can redraw. This study adds the measured share, splits it by an English Wikipedia article, size, listing, region and the company's own Organization markup, and reads the AI Overview on the same search.

One in five sampled software companies got a knowledge panel for their name

Google showed a knowledge panel for the company on 197 of 899 name searches, 21.9% (95% interval 19.3 to 24.7). Another 57 searches, 6.3%, showed a panel our rule did not count as the company's: some about another entity with a similar name (an animal, a film, a city, a local shop), some about the company under another name or without a company-like type line. In all, 28.3% of searches showed some panel.

The panel slot holds two kinds of panel, and we split them after reading the results (found after review). A Knowledge Graph panel describes an entity; a Business Profile panel describes a place, the kind a company claims in Google Business Profile, with a type line such as "Software company in Cardiff, Wales". Of the 899 companies, 15.0% got a Knowledge Graph panel (12.8 to 17.5) and 6.9% a Business Profile panel (5.4 to 8.7).

The share holds under both alternative definitions we fixed before collecting. Counting any panel whose title matches the name, whatever it describes, gives 24.8% (22.1 to 27.7). Requiring a matching title and a company-like description line, with no credit for the company's site ranking on the page, gives 20.5% (18.0 to 23.2). That strict figure is also what the planned definition gave before we fixed how the company's site is read from Google's redirect links, a fix made after the first results were seen (see the methodology).

An English Wikipedia article went with more than four times the rate

Of the 252 sampled companies with an English Wikipedia article, 49.2% got a panel (43.1 to 55.3). Of the 647 without one, 11.3% did (9.1 to 14.0): a gap of 37.9 points (31.3 to 44.4). Every company in the sample has a Wikidata item; the 609 with no Wikipedia article in any language got a panel 10.8% of the time (8.6 to 13.6), about one time in nine.

Split by kind (found after review), the gap is sharper. Companies with an English article got a Knowledge Graph panel 46.8% of the time; those without, 2.6% (17 of 647), a gap of 44.2 points (37.9 to 50.4). Without an English article, most panels were Business Profiles: 8.7% of those companies got one, against 2.4% of companies with an article. Among the 609 with no Wikipedia article in any language, 1.8% got a Knowledge Graph panel and 9.0% a Business Profile panel.

The rate climbs with the number of Wikipedia language editions: 10.8% with none (609 companies), 39.1% with 1 to 4 (230), 66.0% with 5 to 19 (53). The gap stays when listing is held constant: among unlisted companies, 45.7% with an article against 11.1% without (223 and 642 companies). Found after review: among companies whose Wikidata record predates 28 August 2024, the gap is 49.4% against 18.0% (249 and 111 companies), so within older records, the only age group where the comparison is possible, the gap remains.

Figure 2Panel share by number of Wikipedia language editions
Panel share by number of Wikipedia language editions
Value
None (609)10.8%
1 to 4 (230)39.1%
5 to 19 (53)66.0%
20 and over (7)85.7%

Source: LoudFace study of 899 software companies in Wikidata, Google results via DataForSEO, 17 November 202617 Nov 2026

7 companies had 20 or more editions; 6 of them got a panel.

Listed companies, older records and North America got more panels

Listed companies got a panel 70.6% of the time (34 companies) against 20.0% for unlisted ones (865). Companies with 1,000 to 9,999 employees in Wikidata got one 68.4% of the time (19 companies) against 20.9% for the rest, though Wikidata gives an employee count for only 71 of the 899. Companies in North America got one 31.4% of the time (306) against 17.0% elsewhere (593); the searches ran from a United States location, which may favour them. Records created on or after 28 August 2024, 539 of the 899, got a panel 10.0% of the time against 39.7% for older records. All of these gaps stay significant after Holm correction across the ten planned comparisons.

By industry, software-as-a-service companies got one 34.3% of the time (67 companies), computer security 15.7% (102), cloud computing 25.9% (27) and the broad software industry 21.5% (703); the small groups give wide intervals and we did not test them against each other.

Figure 3Panel share by listing, region and age of the Wikidata record
Panel share by listing, region and age of the Wikidata record
Value
Listed (34)70.6%
Not listed (865)20.0%
North America (306)31.4%
Europe (306)12.7%
Asia (96)22.9%
Record before 28 Aug 2024 (360)39.7%
Record from 28 Aug 2024 (539)10.0%

Source: LoudFace study of 899 software companies in Wikidata, Google results via DataForSEO, 17 November 202617 Nov 2026

Organization markup with sameAs did not go with more panels

We read 759 home pages for companies whose search returned. 71.5% carried Organization markup with at least one sameAs link, and 39.5% linked Wikidata or Wikipedia in it. Companies with that markup got a panel 16.4% of the time (543) against 31.5% without it (216). Found after review: much of that reversal goes with record age. Home pages of newer records carried the markup 84.4% of the time against 45.8% for older ones, and within each record age the markup made no clear difference: 9.1% against 15.2% for newer records (427 and 79 companies), 43.1% against 40.9% for older ones (116 and 137).

Figure 4Panel share with and without Organization markup with sameAs, within each record age (found after review)
Panel share with and without Organization markup with sameAs, within each record age (found after review)
Value
Older records, markup (116)43.1%
Older records, no markup (137)40.9%
Newer records, markup (427)9.1%
Newer records, no markup (79)15.2%

Source: LoudFace study of 899 software companies in Wikidata, Google results via DataForSEO, 17 November 202617 Nov 2026

Home pages read only. Found after review: an analysis added after the results were seen.

What the panels show

Of the 197 panels for the company, 49.7% showed a Founded row, 39.6% Headquarters, 18.8% CEO and 18.3% a Date founded row. 13.7% carried a written description.

Figure 5Rows shown in the 197 knowledge panels for the company
Rows shown in the 197 knowledge panels for the company
Value
Founded49.7%
Headquarters39.6%
CEO18.8%
Date founded18.3%
Founder14.7%
Founders12.7%

Source: LoudFace study of 899 software companies in Wikidata, Google results via DataForSEO, 17 November 202617 Nov 2026

An AI Overview appeared on 97.0% of name searches

Google showed an AI Overview on 97.0% of the 899 name searches (95.7 to 97.9). Ahrefs found AI Overviews on 82.91% of US desktop search results for branded keywords on 29 September 2026, and on 73.3% of branded keywords as a September monthly average; our queries are exact company names, many of them small companies, a narrower set than all branded keywords. It cited the company's own site in 59.7% of them (56.5 to 63.0); counting only references that carried a readable address, 49.1%. With a panel for the company on the page, the AI Overview cited the company's site 74.6% of the time (193 searches) against 55.5% without one (679). So on four to five in ten name searches with an AI Overview, depending on how the redirect-link references are read, it cited no page on the company's own site.

What to do about it

  • Search your exact company name from a clean browser. In this sample 21.9% of software companies got a panel and 6.3% saw a panel our rule did not count as theirs, often another entity with a similar name; if that is you, use one consistent name across your site, markup and profiles.
  • Treat Wikipedia as the step that goes with panels, and earn it with coverage that meets its notability rules. Companies with an English article got a panel 49.2% of the time against 11.3% without; never write your own article.
  • Do not count on a Wikidata item alone. Every company here had one, and those with no Wikipedia article in any language got a panel 10.8% of the time; found after review, 1.8% got a Knowledge Graph panel.
  • If your company qualifies for a Google Business Profile, claim and complete it. Found after review: among companies without an English Wikipedia article, Business Profile panels were the more common kind, 8.7% against 2.6% for Knowledge Graph panels.
  • Keep Organization markup for accuracy, not for a panel. 71.5% of the 759 home pages we read had it with sameAs links, and within each record age it made no clear difference (found after review: 43.1% against 40.9% for older records).
  • Read the AI Overview for your name. It appeared on 97.0% of searches and cited the company's own site in 59.7%; make sure the pages it does cite (directories, profiles, review sites) state your facts the way your site does.

Method

How we measured it, and where it stops.

Read this before you quote the numbers. The sheet is the paper’s measurement record: what was sampled, over which window, on which engines, and how.

Study, methodVersion 1.0 · 17 Nov 2026
Sample
900 software companies drawn at random from 3,933 in Wikidata, 899 of them searched by name on Google (United States, desktop) on 17 November 2026, with 760 home pages read
Window
Searched and crawled 17 November 2026
Engines
Google Search, United States, desktop (via DataForSEO)
Published
17 November 2026

Method

Question. When a person in the United States searches a software company's name on Google, how often does Google show a knowledge panel for that company, and what goes with having one? We wrote the method and recorded its hash before collecting any data. Amendment 1 (during collection, before any result was read) uses North America in place of the United States for the region cut, because the frame query read the continent and not the country, and records that a first collection run was stopped by a time limit before any search result was saved. Amendment 2 (after the first results were seen) fixed how the company's site is read from Google's results; it moved the headline from 20.5% to 21.9%, and the strict robustness definition, which it does not touch, stays at 20.5%. Amendment 3 (after the results were seen) added the record-age checks, and Amendment 4 the split between Knowledge Graph and Business Profile panels; the paper labels both as found after review. Amendment 5 (after the second verifier read the paper) brought the count of Wikipedia editions back to the method's definition: the frame query had counted links to every Wikimedia project, Commons included, so the count was re-read for the 900 sampled items from Wikipedia sites only. It moved 41 of the 900 rows and the "no Wikipedia article in any language" group from 599 to 609 companies; the English-article flag did not change. All five are in the published method file.

Wikipedia fields. English Wikipedia article: the Wikidata item links an article on en.wikipedia.org. Wikipedia editions: the number of Wikipedia articles, in any language, linked to the item, counting Wikipedia sites only (Wikidata's site group "wikipedia"), read for the sampled items on 17 November 2026.

Frame. Every company in Wikidata (CC0) whose industry is software industry, software as a service, cloud computing or computer security, with an official website and no dissolution date: 4,178 companies, read on 17 November 2026. Excluded and counted: 94 whose website host is shared with another company in the frame (the duplicate check), 4 on a shared platform's host, and 147 with no English name. 3,933 remained. A company with several websites keeps the shortest address.

Sample. A seeded random shuffle of the 3,933 (seed 20261117) and the first 900 drawn. The method set a target margin of error of ±4 points at 95%, which needs at least 525 companies from a population of 4,178. 899 searches returned; one failed after retries and is left out.

Searching. One Google search per company through DataForSEO's Google Organic SERP Live Advanced endpoint: the company's English name from Wikidata exactly as stored, location United States, language English, desktop, the first page of results, with the AI Overview loaded. Collected between 11:53 and 12:07 UTC on 17 November 2026.

What counts. A name match: the panel's title and the company's name are equal after lower-casing, removing accents and punctuation, reading & as "and" and removing these trailing words: inc, incorporated, corp, corporation, co, company, ltd, limited, llc, plc, gmbh, ag, sa, sas, bv, nv, ab, oy, as, srl, spa, pty, holdings, group; or one starts with the other followed by a space. A panel for the company: Google shows a knowledge panel, its title is a name match, and either its type line contains one of company, corporation, software, business, developer, enterprise, firm, provider, startup, subsidiary, organization (or organisation), vendor, platform, service, website, application or app, or the company's own site is among the first-page results. The headline is the share of the 899 returned searches with a panel for the company. A panel about another entity is any other panel. The AI Overview cites the company when one of its references is on the company's host or a subdomain, or, for references Google returns as a redirect link with only a site name, when that name matches the company's name or its host. Organization markup with sameAs: a JSON-LD node on the home page whose type is Organization or a subtype, with at least one sameAs value. Found after review: a panel that carries a Google place id is a Business Profile panel; one without is a Knowledge Graph panel.

Home pages. Our crawler (LoudFaceResearchBot) read robots.txt first and the home page only where robots.txt allowed it, one request per second per host, 15-second timeout. 760 of 900 home pages were read, 759 of them for companies whose search returned; the markup shares use those 759.

Analysis. Shares with Wilson 95% intervals; gaps between two groups with Newcombe intervals; ten planned comparisons (Wikipedia article, listing, each employee band against the rest, record age, North America against the rest, markup) with two-proportion tests and Holm correction. Robustness: the headline and the Wikipedia gap re-run under a looser definition (any panel with a matching title) and a stricter one (a matching title and a company-like type line, no credit for the company's site ranking).

Data checks. Duplicate checks: hosts shared by several companies were excluded (94) and no Wikidata item appears twice. Bot checks: there is no visitor data, so no bot traffic can enter the study; 43 home pages answered our declared crawler with a block (an HTTP 401 or 403 or 429 status) and count as not read. Outlier checks: no employee count above 1,000,000 and no company name longer than 60 characters, so 0 outliers. 5 home pages had JSON-LD that would not parse; their valid blocks were still read.

Reproduce it. The frame query, the draw, the collector and the analysis are in the published files; the per-company CSV holds every input the shares are computed from.

Limitations

Coverage. 899 of 900 Google searches returned. Home pages: 760 of 900 read (84.4%); 84 sites could not be reached or returned a server error when we asked for robots.txt, 43 blocked our declared crawler, 8 disallowed it in robots.txt and 5 failed otherwise. The markup figures cover the 759 of those home pages whose company's search returned. Companies whose home page we could not read got a panel 28.6% of the time, so the markup comparison leaves out a group that differs from the rest.

The frame. Wikidata lists companies notable or keen enough to be added, and 539 of the 899 have records created since 28 August 2024, and found after review, 3 of those 539 have an English Wikipedia article. This is an open, re-runnable sample of software companies in Wikidata, not a census of software companies, and every company in it has a Wikidata item, which most companies do not.

Measurement. One search per company, from one United States location, signed out, on one day. Results vary by place, person and time, and a panel can come and go. The query is the name as Wikidata stores it; a person may type a shorter or longer name. The name-match rule can miss a panel whose title differs from Wikidata's label, and the type-line rule can let in a panel about a product with the company's name. The two robustness definitions show how far those choices move the headline: 20.5% and 24.8%. The headline counts Business Profile panels, which a company can claim itself, alongside Knowledge Graph panels; the split is reported, found after review, and it reads a Google place id on the panel as the mark of a Business Profile, a proxy that can misfile a panel. Some Google links came back as redirect links with only a site name, so the AI Overview's own-site share uses that name for those references; counting only references with a readable address gives a lower bound of 49.1%.

Size and region. Wikidata gives an employee count for 71 of the 899, so the size cut rests on small groups. The region cut uses the continent, so North America includes Canada and Mexico, and the searches ran from the United States, so the region gap mixes where a company is with where we searched from.

Cause. Every comparison is an association. Companies with a Wikipedia article are older, better known and more often listed; the gap holds within unlisted companies and, found after review, within older records, but notability can drive both the article and the panel.

Versions

  • Version 1.0 · 17 Nov 2026CurrentFirst edition: 899 software companies searched by name on 17 November 2026.
  • Version 1.1When the paper is re-run, the new version is listed here and this page links to it.

Data and citation

  • Per-company data (CSV, one row per sampled company)Company names and Wikidata fields (CC0); whether the search returned, the panel's title and type line, whether it counts as a panel for the company under each definition, the panel's rows, the AI Overview and whether it cited the company's site, and the home page's markup.

Bukva, A. (2026, November 17). How often Google shows a knowledge panel for a software company's name, and what goes with one: 899 companies from Wikidata (2026). LoudFace Research. https://www.loudface.co/research/software-company-knowledge-panels

APA

Questions

What readers ask about this paper

How many software companies have a Google knowledge panel?
In this study, 21.9% of 899 software companies drawn at random from Wikidata got a knowledge panel for the company when their name was searched on Google from the United States on 17 November 2026 (95% interval 19.3 to 24.7). Another 6.3% saw a panel our rule did not count as the company's, often about another entity with a similar name.
Does a Wikipedia article help a company get a knowledge panel?
It goes with one. Companies with an English Wikipedia article got a panel 49.2% of the time against 11.3% for those without, and the gap held among unlisted companies (45.7% against 11.1%). The data shows an association; notability can drive both the article and the panel.
Is a Wikidata entry enough for a knowledge panel?
Not on its own in this sample. Every company studied has a Wikidata item, and those with no Wikipedia article in any language got a panel 10.8% of the time. Found after review: most of those were Business Profile panels, and 1.8% got a Knowledge Graph panel.
Does Organization schema markup get you a knowledge panel?
We found no sign of it. Home pages with Organization markup and sameAs links got a panel 16.4% of the time against 31.5% without, and, found after review, within each record age the difference was not clear (43.1% against 40.9% for older records).
Does Google show an AI Overview when you search a company's name?
On 97.0% of the 899 name searches in this study. It cited the company's own site in 59.7% of them, and in 74.6% when a knowledge panel for the company was also on the page.

More research

Keep reading. Each cover is the finding.

All papers →
AI search behaviourStudy · 10 Nov 2026

gpt-6.1-sol with search gave 13.7% of 444 software firms a founding year unlike Wikidata's

How often each setting gave a founding year that differs from Wikidata's
ValueNote
gpt-6.1-sol with web search13.7%10.8 to 17.3
gpt-6.1-sol from memory9.9%7.5 to 13.0
gpt-5.4-mini with web search14.9%11.9 to 18.5
gpt-5.4-mini from memory16.4%13.3 to 20.2
444 software companies drawn at random from 2,757 in Wikidata, each asked once in four settings (gpt-6.1-sol and gpt-5.4-mini, from memory and with web search), 3 November 2026Read →
AI Visibility IndexIndex · 30 Jul 2026

The same fintech brand can be 4.6 times more visible on one engine

Visibility per engine
ChatGPTAI OverviewsPerplexity
Remote62.61%48.9%49.44%
Deel61.48%60.61%36.75%
Papaya Global34.63%15.31%22.66%
Our client15.67%36.19%7.92%
Bitwage10.4%14.01%5.55%
23,294 conversationsRead →
State of B2B SaaSReport · 3 Nov 2026

One in twelve software companies in Wikidata blocks an AI crawler by name

What software companies block in robots.txt
Value
Block a training crawler by name8.4%
Block every crawler through * and name no AI crawler0.6%
Block a search or user crawler by name0.2%
3,263 software companies in Wikidata with a readable robots.txt (of 3,884 sites tried), 15 AI crawlers, crawled on 3 November 2026Read →
A 30-minute strategy call

Find out where AI answers
put you today.

We run this measurement for your brand: which sources ChatGPT, Perplexity and Google AI Overviews cite in your category, where your domain ranks, and what we would do first, with a written scope within 48 hours.

Prefer email? Write to contact@loudface.co
We reply within two hours during working hours