Tool AT-W15REV. 2026-A
Cookie banner needed? Check Google Fonts and third-party services
A website needs a cookie banner only if it embeds services that require consent: analytics with cookies, advertising pixels, videos and maps that load immediately. A court took the same view of Google Fonts loaded from Google's server. The check fetches your page, names every third-party server with where it was found, and tells you how to manage without a banner.
- free
- no sign-up
- 5-minute cache
01 / PAGE FETCH
Enter an address, see the third-party services
Page fetch · third-party services and cookies
The check reads the source code of a page and classifies the services it finds. This is a technical assessment, not legal advice; if in doubt, a lawyer can say what applies to your page.
02 / HOW IT WORKS
How does the third-party check work?
You enter an address and our server fetches the page once, like a browser that does not run JavaScript. Your own browser talks only to aureum-tech.com. The server reads the source code and looks for every place where the page fetches something from a third-party server. It also opens up to eight of the page's stylesheets, because Google fonts are often hidden in a CSS file rather than in the HTML.
The check compares each server it finds with a catalogue of 59 services. It also recognises services from the script code, for example the call gtag() from Google Analytics or fbq() from the Meta Pixel. In addition it reads which cookies the server sets directly and whether a consent tool such as Cookiebot, Usercentrics or Borlabs is installed. Servers that the catalogue does not know are shown anyway.
What matters is when something loads. A script that starts immediately on opening counts differently from one that a consent tool holds back until the visitor agrees. This produces the overall picture: banner probably needed, unclear, or probably not needed. A banner that does not hold the services back is reported separately. That is not rare: in one study, 49% of websites set profiling cookies before anyone had consented Trevisan 2019.
03 / BACKGROUND
When does a website need a cookie banner?
The name is misleading. It is not about cookies as such but about two rules. The first is in § 25 TDDDG (until May 2024 the law was called TTDSG): anyone who stores information on the visitor's device or reads it there needs the visitor's consent. The exception is what is strictly necessary for the visitor to receive the service they expressly requested, such as a shopping basket or a login. For analytics and advertising this is hard to justify; Microsoft itself describes the cookies of its analytics service Clarity as not necessary.
The second rule is the GDPR (DSGVO). Every request to a third-party server transmits the visitor's IP address, which counts as personal data. That needs a legal basis. In 2022 the Regional Court of Munich I ruled that a website was not allowed to pass a visitor's IP address to Google via Google Fonts: a legitimate interest was lacking because the font can also be used without a connection to Google. This argument applies to any file that could just as well come from your own server.
Two judgments of the European Court of Justice belong here. Under Planet49, a pre-ticked box is not consent, even if no personal data is stored. Under Fashion ID, the operator of a website that embeds a Facebook button shares responsibility for collecting and transmitting the data. Anyone who embeds third-party services therefore cannot shift the responsibility onto the provider.
In practice this means: the banner is not an obligation that applies to every website. It is a consequence of the technology embedded. Without your consent, this website makes no request to third-party servers except Cloudflare, which delivers it; it asks only on the appointment page, because a Google calendar loads there. The guide Website without a cookie banner explains the approach in detail. What is written here is a technical assessment and not legal advice.
Sources: § 25 TDDDG (text of the law) · LG München I, 3 O 17493/20 · ECJ C-673/17 · ECJ C-40/17
04 / SERVICES
Which services need consent
The table summarises the check's catalogue by type of service. Brackets show where an individual service departs from the rule for its group.
| Type | Examples | Banner question | What works instead |
|---|---|---|---|
| Fonts | Google Fonts (consent needed as usually embedded), Adobe Fonts (Typekit), Font Awesome | no cookie needed, but IP address goes to a third party | Download the font files and deliver them from your own server. The page looks exactly the same to visitors afterwards. |
| Analytics | Google Tag Manager, Google Analytics, Matomo Cloud, Matomo on your own or a commissioned server, Hotjar, Microsoft Clarity | Consent needed as usually embedded | Count on your own server without a script in the browser, or run a cookie-free analytics program on your own server. This is how this website does it. |
| Maps | Google Maps, OpenStreetMap (no cookie needed, but IP address goes to a third party) | Consent needed as usually embedded | A map section as an image from your own server and a link to the route planner. Or load the map only after a click (two-click solution). |
| Video | YouTube in enhanced privacy mode (provider calls its cookie necessary, not settled), YouTube preview image (no cookie needed, but IP address goes to a third party), YouTube, Vimeo | Consent needed as usually embedded | A preview image from your own server with a note that data goes to the provider when the video plays; the video loads only after the click. Short films can also be delivered from your own server. |
| Social networks | Instagram embed, Facebook plugins and embeds, X (Twitter) | Consent needed as usually embedded | A simple link to the profile, with its own icon from your own server. Posts as a screenshot with a link, or loaded only after a click. |
| Form protection | Google reCAPTCHA, hCaptcha, Cloudflare Turnstile | Provider calls its cookie necessary, not settled | An invisible extra field (honeypot), a timing check and a rate limit on your own server keep many spam bots out, with no third-party server. |
| Advertising and pixels | Google Ads and advertising network, Meta Pixel (Facebook), LinkedIn Insight Tag, TikTok Pixel, Microsoft Advertising (UET) | Consent needed as usually embedded | Leave out, or load only after consent. You can measure the success of ads without a pixel too, for example with your own landing page address for each campaign. |
| Chat and widgets | Calendly, Trustpilot widget (no cookie needed, but IP address goes to a third party), ProvenExpert seal (no cookie needed, but IP address goes to a third party), tawk.to chat, Intercom chat, Zendesk widget | Consent needed as usually embedded | A button or link that loads the widget only after a click, or opens the service in a new window. |
| Third-party libraries (CDN) | Google Hosted Libraries, cdnjs, jsDelivr, unpkg, jQuery CDN, BootstrapCDN | no cookie needed, but IP address goes to a third party | Download the file, bring it up to date while you do so, and deliver it from your own server. Usually only one line changes in the source code, and the page no longer depends on the third-party server. |
| Consent dialogue | Cookiebot, Usercentrics, OneTrust, Borlabs Cookie, Complianz, Real Cookie Banner | no banner reason of its own | Check which services are entered in it. If nothing that needs consent remains after the rebuild, the dialogue can go. |
| Website builder and shop system | Wix, Jimdo, Shopify, Squarespace, Webflow | no banner reason of its own | Conclude the data processing agreement in the provider's account and name it in the privacy policy. |
The table describes the normal case. What matters is how a service is embedded. A video that loads only after a click, with a short notice, obtains consent with that click. If the same video loads immediately when the page is opened, it needs a banner. That is why the check shows for each finding whether it loads immediately, with a delay, or only after consent.
05 / REBUILD
How your website gets by without a banner
- Fonts onto your own server. Download Google Fonts, icon fonts such as Font Awesome and Adobe fonts with a suitable licence and deliver them yourself. Then search the stylesheets for
fonts.googleapis.com, because themes like to embed fonts a second time. - Fetch libraries from the CDN. Store jQuery, Bootstrap and the like as a file and bring them up to date while you do so. In one study, more than a third of websites embedded a library with a known vulnerability Lauinger 2017.
- Analytics without cookies. Count page views on your own server, or run Matomo yourself and switch the cookies off. Google Analytics, Hotjar and Clarity need consent.
- Maps and videos only after a click. A preview image from your own server, with a sentence underneath saying that data goes to the provider when it plays. For directions, a map image with a link to the route planner is often enough.
- Form protection without a third-party server. An invisible extra field and a timing check on your own server instead of reCAPTCHA, which loads on every page.
- Clear out pixels and tag managers. Whatever is to stay loads only after consent. Whatever nobody evaluates any more comes out.
- Check again, then decide about the banner. If the check no longer shows a service that needs consent, the banner can go. The privacy policy then describes what the page actually does.
What a policy looks like when hardly any third-party services are left is shown by our own privacy policy (Datenschutzerklärung). If the page runs on a website builder or shop system, its provider is your service provider; for that you need a data processing agreement (Vertrag zur Auftragsverarbeitung). We carry out the rebuild as part of web design or as a one-off job on an existing page. A side effect: in one study, third-party services lengthened load time by up to 50% in the extreme case Goel 2017.
06 / LIMITS
What the check does not see, and how to look for yourself
- Only the page entered is checked, not subpages. The check sees a video on the contact page only if you enter the contact page.
- JavaScript is not run. Anything a script loads only in the browser, for example via Google Tag Manager, the check sees only if the address appears in the source code.
- The check sees cookies only if the server sets them directly. Most tracking cookies are created by scripts in the browser.
- Some servers deliver different content to a checking program than to a browser, or block it.
- The legal classification is a technical assessment under § 25 TDDDG and the GDPR (DSGVO), not legal advice.
The first limit in particular matters. About half of all websites show content that they did not load themselves but that an embedded third party did Ikram 2019. A green result therefore means: nothing in the source code requires consent. It is not a guarantee for everything that happens in the browser.
Looking for yourself in the developer tools
- Open the page in Chrome, Edge or Firefox and press F12 (Mac: Option + Command + I).
- Choose the “Network” tab and tick “Disable cache” there.
- Reload the page without clicking anything in the banner.
- Show the “Domain” column (right-click the column headings) and sort by it. Every row with a third-party domain is a request to a third party.
- Under “Application” or “Storage” you will find the page's cookies and browser storage.
This check looks only at third-party services and answers at once, with no e-mail address. How well your page is found on Google and in AI assistants is checked by SpotLens with a detailed report.
07 / RESEARCH
What the research says
Cookie banners and embedded third-party services have been measured thoroughly. The studies show above all how wide the gap is between “banner present” and “consent valid”. Six findings that went into the check:
- A lot loads before the decision. 49% of the websites measured set profiling cookies before consent Trevisan 2019. In a later study, over 75% of tracking events took place before visitors could choose or after they had declined Papadogiannakis 2021. That is why the check assesses the first visit.
- A banner is not yet consent. Only 11.8% of the dialogues examined met minimum requirements of European law Nouwens 2020, and for a widespread banner standard, 54% of the pages examined more closely showed at least one suspected violation Matte 2020.
- Design steers the answer. Field experiments with over 80,000 visitors showed large differences depending on position and preselection Utz 2019; when “Decline” was missing on the first level, agreement rose by 22 to 23 percentage points Nouwens 2020. Anyone with no services that need consent does not have to ask these questions at all.
- Tracking does not need cookies. Besides cookies, fingerprinting and matching of identifiers are used Englehardt 2016. That is why the check also looks for traces in the script code.
- No catalogue is complete. Common filter lists missed 25 to 30% of the trackers that a study on invisible tracking pixels found Fouad 2020. The check therefore shows unknown third-party servers with where they were found.
- Third-party code is a risk. Over 90% of websites loaded content from third-party servers Kumar 2017, and 37% embedded at least one library with a known vulnerability Lauinger 2017. Delivering it yourself solves the banner question and this problem in one go.
08 / FOR AI ASSISTANTS
Using this tool from an AI assistant
Assistants such as Claude or ChatGPT that speak the Model Context Protocol (MCP) can call this page’s calculation directly instead of rebuilding it. The server is public, needs no sign-in and stores nothing. The answer always includes the link to this page, the date of the data and the limits of the result.
- Endpoint
https://aureum-tech.com/mcp- Tool
werkzeug_drittanbieter_check- Transport
- Streamable HTTP, read-only
| Name | Required | Meaning |
|---|---|---|
url | yes | Address of the web page, e.g. your-company.de or https://www.your-company.de/contact. Without http://, https:// is assumed. Public websites only, port 80 or 443. |
Questions where an assistant should call it
- “Does my website handwerk-muster.de need a cookie banner?”
- “Does www.beispiel-praxis.de load Google Fonts from Google servers?”
- “Which third-party services does our club's home page embed?”
- “Our banner annoys visitors. Can it go?”
How the connection works and what MCP can do in your own business is explained on the page about MCP integrations.
09 / COMMON QUESTIONS
Common questions about cookie banners and third-party services
Does every website need a cookie banner?
No. A website needs consent if it stores or reads information in the browser that is not strictly necessary for the service requested (§ 25 TDDDG), or if it passes data to third parties without another legal basis. A shopping basket and a login need none. A page with its own fonts, no tracking and no videos that load immediately gets by without a banner.
May I use Google Fonts?
Yes, if the fonts come from your own server. If the page loads them from Google when it is opened, every visitor's IP address goes to Google. The Regional Court of Munich I (LG München I) held this unlawful without consent on 20.01.2022 and awarded 100 euros in damages (3 O 17493/20). Google's fonts are freely licensed and may be stored on your own server.
Is YouTube in enhanced privacy mode (youtube-nocookie.com) enough?
It helps, but it does not settle the question. All Google says about the mode is that views do not affect what the visitor later sees on YouTube and that the advertising is not personalised. The player still loads from Google servers when the page is opened, so the IP address goes there. A preview image from your own server is safe, with the video loading only after a click.
I have a banner. Why does the check still report red?
Because services load on the first visit, before anyone has chosen anything in the banner. The banner then does not hold them back. This is not a one-off: in one study, over 75% of tracking events took place before visitors could choose or after they had declined. The remedy is to have the consent tool block the services, or to remove them altogether.
Why does the check miss a service that I can see in the browser?
The check reads the source code of a single page and does not run JavaScript. Anything a script loads only in the browser, for example via Google Tag Manager, it sees only if the address appears in the source code. About half of all websites show content that an embedded third party has loaded. You check services on subpages by entering their addresses.
Do I need a banner for Matomo?
With cookies, yes, under § 25 TDDDG. According to its maker, however, Matomo can run without cookies, using _paq.push(['disableCookies']) before the tracking call. If it also runs on your own server, the data goes to nobody else. The measurement still belongs in the privacy policy (Datenschutzerklärung). The check recognises from the script code whether cookies are switched off.
What happens to the address I enter?
It goes to our server, which fetches the page once like a browser without JavaScript, plus up to eight stylesheets. The address and result are excluded from our logs. Both stay in memory for up to five minutes for repeat checks. To prevent misuse, the server counts requests per visitor as an anonymous check value. It does not fetch internal addresses.
10 / STUDIES AND SOURCES
What the tool is based on
The benchmarks and notes above rest on these 14 scientific papers. Each was checked through its DOI or the publisher's page (last checked on 13 November 2026): it exists, it has not been retracted, and its finding still holds for the question today.
Trevisan, M.; Traverso, S.; Bassi, E.; Mellia, M. (2019): 4 Years of EU Cookie Law: Results and Lessons Learned. Proceedings on Privacy Enhancing Technologies 2019 (2), pp. 126–145
Finding In a large measurement of European websites, 49% did not comply with the cookie policy: they set profiling cookies before the visitor had consented.
In the tool That is why the check distinguishes whether a service loads immediately or waits for consent, and reports “loads before consent” separately.
doi.org/10.2478/popets-2019-0023 · Journal article
Papadogiannakis, E.; Papadopoulos, P.; Kourtellis, N.; Markatos, E. P. (2021): User Tracking in the Post-cookie Era: How Websites Bypass GDPR Consent to Track Users. Proceedings of the Web Conference 2021 (WWW), pp. 2130–2141
Finding More than 75% of the tracking events measured took place before visitors could choose in the banner, or after they had declined everything. Besides cookies, ID sharing and fingerprinting were used.
In the tool The check therefore assesses what loads on the first visit, not whether a banner is there.
doi.org/10.1145/3442381.3450056 · Conference paper
Matte, C.; Bielova, N.; Santos, C. (2020): Do Cookie Banners Respect my Choice? Measuring Legal Compliance of Banners from IAB Europe's Transparency and Consent Framework. 2020 IEEE Symposium on Security and Privacy (S&P), pp. 791–809
Finding Of 1,426 websites using a widespread banner standard, 141 stored a consent before the visitor had chosen anything. Of 560 examined more closely, 54% showed at least one suspected violation.
In the tool A consent dialogue that is detected does not count as all-clear in the check; what matters is what it actually holds back.
doi.org/10.1109/SP40000.2020.00076 · Conference paper
Nouwens, M.; Liccardi, I.; Veale, M. et al. (2020): Dark Patterns after the GDPR: Scraping Consent Pop-ups and Demonstrating their Influence. Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems, pp. 1–13
Finding Only 11.8% of the consent dialogues examined on British websites met minimum requirements of European law. When “Decline” was missing on the first level, agreement rose by 22 to 23 percentage points.
In the tool The text advises rebuilding the services first; where a dialogue remains, declining belongs on the first level with equal weight.
doi.org/10.1145/3313831.3376321 · Conference paper
Utz, C.; Degeling, M.; Fahl, S.; Schaub, F.; Holz, T. (2019): (Un)informed Consent: Studying GDPR Consent Notices in the Field. Proceedings of the 2019 ACM SIGSAC Conference on Computer and Communications Security (CCS), pp. 973–990
Finding Field experiments with over 80,000 visitors to a German website: position, type of choice and preset ticks change the decision strongly. Over 60% of popular European websites showed consent notices.
In the tool Shows why a page without services that need consent does better than one with the best-designed banner.
doi.org/10.1145/3319535.3354212 · Conference paper
Degeling, M.; Utz, C.; Lentzsch, C. et al. (2019): We Value Your Privacy ... Now Take Some Cookies: Measuring the GDPR's Impact on Web Privacy. Network and Distributed System Security Symposium (NDSS) 2019
Finding After the GDPR (DSGVO) came into force, 62.1% of the European websites examined showed cookie notices. Technically, switching off third-party cookies depends on the third party cooperating.
In the tool That is why the check recommends leaving out the third-party service or delivering it yourself, rather than merely managing it with a banner.
doi.org/10.14722/ndss.2019.23378 · Conference paper
Sanchez-Rola, I.; Dell'Amico, M.; Kotzias, P. et al. (2019): Can I Opt Out Yet? GDPR and the Global Illusion of Cookie Control. Proceedings of the 2019 ACM Asia Conference on Computer and Communications Security (AsiaCCS), pp. 340–351
Finding On over 90% of 2,000 high-traffic websites there were cookies that can be used to recognise visitors; many pages made misleading statements about them.
In the tool The check lists the cookies that the server sets directly, with their lifetime, rather than relying on the statements in the banner.
doi.org/10.1145/3321705.3329806 · Conference paper
Englehardt, S.; Narayanan, A. (2016): Online Tracking: A 1-million-site Measurement and Analysis. Proceedings of the 2016 ACM SIGSAC Conference on Computer and Communications Security (CCS), pp. 1388–1401
Finding A measurement of the one million most-visited websites found tracking with cookies and without them (fingerprinting), including methods never before measured in use, and matching of identifiers between providers.
In the tool That is why the check also looks for signatures such as gtag() or fbq() in the script code, not only for cookies.
doi.org/10.1145/2976749.2978313 · Conference paper
Fouad, I.; Bielova, N.; Legout, A.; Sarafijanovic-Djukic, N. (2020): Missed by Filter Lists: Detecting Unknown Third-Party Trackers with Invisible Pixels. Proceedings on Privacy Enhancing Technologies 2020 (2), pp. 499–518
Finding Invisible tracking pixels from third-party servers were found on over 94% of the domains examined. Common filter lists missed 25 to 30% of the trackers that the study identified.
In the tool The catalogue cannot be complete; that is why the check shows every unknown third-party server with where it was found, instead of concealing it.
doi.org/10.2478/popets-2020-0038 · Journal article
Ikram, M.; Masood, R.; Tyson, G. et al. (2019): The Chain of Implicit Trust: An Analysis of the Web Third-party Resources Loading. The World Wide Web Conference (WWW) 2019, pp. 2851–2857
Finding About half of all websites showed content that they had not loaded themselves but that an embedded third party loaded afterwards; some chains went beyond 30 levels.
In the tool Explains the limit of the check: what a third-party script loads in the browser is not in the source code and needs a look in the developer tools.
doi.org/10.1145/3308558.3313521 · Conference paper
Kumar, D.; Ma, Z.; Durumeric, Z. et al. (2017): Security Challenges in an Increasingly Tangled Web. Proceedings of the 26th International Conference on World Wide Web (WWW), pp. 677–684
Finding Over 90% of the websites examined loaded content from third-party servers in 2016; every such dependency enlarges the attack surface, because the third-party content is trusted without being checked.
In the tool Supports the advice to deliver libraries and fonts from your own server.
doi.org/10.1145/3038912.3052686 · Conference paper
Lauinger, T.; Chaabane, A.; Arshad, S. et al. (2017): Thou Shalt Not Depend on Me: Analysing the Use of Outdated JavaScript Libraries on the Web. Network and Distributed System Security Symposium (NDSS) 2017
Finding 37% of over 133,000 websites embedded at least one library with a known security vulnerability; libraries that arrived through advertising or tracking code were affected more often.
In the tool Anyone who moves a library from a CDN onto their own server should bring it up to date at the same time; the check says so under the alternative.
doi.org/10.14722/ndss.2017.23414 · Conference paper
Goel, U.; Steiner, M.; Wittie, M. P.; Flack, M.; Ludin, S. (2017): Measuring What is Not Ours: A Tale of 3rd Party Performance. Passive and Active Measurement (PAM 2017), Lecture Notes in Computer Science, pp. 142–155
Finding Third-party services lengthened the load time of web pages by up to 50% in the extreme case. When their files were delivered from the site's own infrastructure, the pages loaded up to 25% faster.
In the tool Shows the side effect of the “deliver it yourself” alternative: fewer third-party servers, and often faster pages too.
doi.org/10.1007/978-3-319-54328-4_11 · Conference paper
Hils, M.; Woods, D. W.; Böhme, R. (2020): Measuring the Emergence of Consent Management on the Web. Proceedings of the ACM Internet Measurement Conference (IMC) 2020, pp. 317–332
Finding The use of ready-made consent tools doubled from June 2018 to June 2019 and doubled again by June 2020; anyone who wanted to decline a lot needed noticeably longer in the dialogues.
In the tool The check recognises common consent tools and says when one is installed even though there is nothing in the source code for it to manage.
doi.org/10.1145/3419394.3423647 · Conference paper
Standards, authorities and manufacturer data
- § 25 TDDDG (text of the law) · Federal Ministry of Justice
- LG München I, judgment of 20.01.2022, 3 O 17493/20 · Bayern.Recht
- ECJ, judgment of 01.10.2019, C-673/17 (Planet49), press release · Court of Justice of the European Union
- ECJ, judgment of 29.07.2019, C-40/17 (Fashion ID), press release · Court of Justice of the European Union
- Google: Google Analytics cookies · Google
- Google: reCAPTCHA FAQ · Google
- YouTube: Enhanced privacy mode · Google
- Microsoft: Clarity cookies · Microsoft
- Meta: cookies _fbp and _fbc · Meta
- Vimeo: player parameters · Vimeo
- Matomo: tracking without cookies · Matomo
- Cloudflare: cookies · Cloudflare
11 / MORE TOOLS
Still have a question?
If you are touching the website anyway, it is best to settle two more questions at the same time: does it have to be accessible? The BFSG check answers that. And do e-mails from your domain arrive? The e-mail check is for that.
- AT-W2 · Home & business
Windows 10 check
Does this PC still get security updates, and what needs doing by October 2027? - AT-W3 · Home & business
Wi-Fi planner
How many access points or mesh units does your home, office or warehouse need? - AT-W4 · Businesses
IT emergency check
How quickly could your business get back to work after an outage? - AT-W5 · Businesses
E-mail check
Are your e-mails landing in spam? Your domain’s SPF, DKIM and DMARC in plain language. - AT-W6 · Businesses
UPS calculator
Which UPS keeps your server, router and till running long enough during a power cut? - AT-W8 · Businesses
XRechnung viewer
Received an XRechnung as an XML file? This makes it readable without the file leaving your computer.
Would you like to know whether your banner can go without replacement? We will look at the page with you, move fonts and libraries onto your own server, and rebuild videos and maps so that they load only after a click.
+49 3744 4399760[email protected]Mon–Fri 8:00–17:00 · Straße der Jugend 5, 08228 Rodewisch