Audit of digital brand identity: where to start
Which element to start with: website, markup, social networks, or standards
You should start with the website—but not its design, but its technical foundation: the Organization block in JSON-LD. This is the primary source from which search engines, directories, and language models extract facts about the company: name, logo, address, and links to profiles. If this block is empty, incomplete, or contains outdated data, all other channels will inherit the error.
The logic is as follows: first, internal brand signals are checked (website and its markup), then external confirmations (directories, media, reviews), and only after that—perception on user channels, including social networks and AI platforms. Internal brandbook standards come last in the queue, because they regulate how the brand looks, not what systems and algorithms know about it.
Three-layer model: from markup to AI perception
Classic brand audit is 6-7 steps, from goal-setting to final recommendations. Such a framework works for visual identity, but does not cover the digital footprint entirely. It is more practical to structure the check in three layers.
- Own signals — Organization and sameAs markup on the website, meta tags, domain and subdomain structure, SSL certificates.
- External confirmations — profiles in Google Business Profile, Yandex Business, industry directories, mentions in media, Wikipedia profile.
- Real perception by models — what ChatGPT, Gemini, and Perplexity respond to direct questions about the brand.
Each subsequent layer is built on the previous one. If the site markup contradicts the data in the directory, the model will most likely choose the source with greater authority—and this will not always be your site.
Consistency across channels: what and how to check
Consistency is not about using the same font in Instagram and on the website. It's about fact matching: company names, addresses, phone numbers, working hours, legal names. The verification is based on specific points of comparison.
| Channel | What We Test | Typical discrepancy |
|---|---|---|
| Website (JSON-LD) | name, sameAs, logo, address | Outdated logo, missing links to profiles |
| Directories (Google, Yandex) | Name, category, phone number | Old name after rebranding |
| Social Media | Bio, website link, visual style | Link leads to an old domain |
| Media and mentions | Correctness of brand citation | Spelling with hyphen/without hyphen differs |
According to 'Ashmanov and Partners', 73% of consumers consider company reputation an important factor in decision-making, 31% rely on the website, and 20% on reviews and media publications. Discrepancies in these sources directly impact conversion, not just 'brand aesthetics'.
Case: how an invisible markup error destroys the digital footprint
A regional network of clinics conducted a rebranding: it changed its name, logo, and updated its website. Visually everything was perfect — a new corporate style, fresh design, and neat social media profiles. However, the JSON-LD Organization block still contained the sameAs field with old links to profiles that were still listed under the former name in three industry directories.
Result: Google Business Profile continued to display the old name in the knowledge panel, and when users asked ChatGPT "which clinic is better in the city," the model would name the company by its old name — because the training data and directories had not yet caught up with the rebranding. A routine visual audit would not have detected this issue: the website looked flawless. The problem only surfaced when checking the chain "markup → directory → AI response".
How to check whether brands are correctly recognized by ChatGPT and Gemini
Yes, this is part of a digital identity audit, not a separate discipline — it's just rarely included in classic checklists. The mechanism is simple: a set of 20 control prompts is compiled, arranged by stages of decision-making — from the general category request to comparison with specific competitors. Then two metrics are calculated.
- Share of Model — the share of prompts in which the model mentions the brand at all. A direct analog of Share of Voice from media planning.
- Citation Rate — how often the model confirms the answer by referencing the source. Here platforms differ significantly: Perplexity positions its citation accuracy at 97%, but independent audits show significant discrepancies between the source and the cited fact. The citation mechanism varies greatly between platforms: some embed links in most answers, others mention sources less frequently.
A low Citation Rate for ChatGPT does not mean the brand is not visible there — it means you need to directly verify the mention of the brand through prompts, rather than relying on transparent links. Tools like KlevelBrand formalize this into a Brand Score based on 10 parameters, evaluating consistency and visibility in a single figure from 0 to 100.
Minimal checklist for small business
A full three-layer audit requires resources that small businesses often lack. However, it is possible to conduct a quick diagnosis in one day if you limit yourself to priority points.
- Check the Organization block in JSON-LD: whether name, logo, and sameAs with valid links are present.
- Verify the name, address, and phone number on the website against the listings in Google and Yandex directories.
- Ensure that links from social networks lead to the relevant domain without redirects through old pages.
- Ask 5-10 direct questions about the brand in ChatGPT and Gemini and record whether the company is mentioned and what exactly is said.
- Check that most website visitors study it before purchasing and see the current design — data shows that 75% of people evaluate the reliability of a business based on the quality of the website design.
This set is sufficient to identify critical gaps between how the brand appears inside the company and what people and algorithms see from the outside. A full audit with Brand Score calculation and monitoring of Share of Model makes sense to add when the business expands beyond a single region or begins to seriously compete for visibility in AI search.
Which element to start with: website, markup, social networks, or standards
You should start with the website—but not its design, but its technical foundation: the Organization block in JSON-LD. This is the primary source from which search engines, directories, and language models extract facts about the company: name, logo, address, and links to profiles. If this block is empty, incomplete, or contains outdated data, all other channels will inherit the error.
The logic is as follows: first, internal brand signals are checked (website and its markup), then external confirmations (directories, media, reviews), and only after that—perception on user channels, including social networks and AI platforms. Internal brandbook standards come last in the queue, because they regulate how the brand looks, not what systems and algorithms know about it.
Three-layer model: from markup to AI perception
Classic brand audit is 6-7 steps, from goal-setting to final recommendations. Such a framework works for visual identity, but does not cover the digital footprint entirely. It is more practical to structure the check in three layers.
- Own signals — Organization and sameAs markup on the website, meta tags, domain and subdomain structure, SSL certificates.
- External confirmations — profiles in Google Business Profile, Yandex Business, industry directories, mentions in media, Wikipedia profile.
- Real perception by models — what ChatGPT, Gemini, and Perplexity respond to direct questions about the brand.
Each subsequent layer is built on the previous one. If the site markup contradicts the data in the directory, the model will most likely choose the source with greater authority—and this will not always be your site.
Consistency across channels: what and how to check
Consistency is not about using the same font in Instagram and on the website. It's about fact matching: company names, addresses, phone numbers, working hours, legal names. The verification is based on specific points of comparison.
| Channel | What We Test | Typical discrepancy |
|---|---|---|
| Website (JSON-LD) | name, sameAs, logo, address | Outdated logo, missing links to profiles |
| Directories (Google, Yandex) | Name, category, phone number | Old name after rebranding |
| Social Media | Bio, website link, visual style | Link leads to an old domain |
| Media and mentions | Correctness of brand citation | Spelling with hyphen/without hyphen differs |
According to 'Ashmanov and Partners', 73% of consumers consider company reputation an important factor in decision-making, 31% rely on the website, and 20% on reviews and media publications. Discrepancies in these sources directly impact conversion, not just 'brand aesthetics'.
Case: how an invisible markup error destroys the digital footprint
A regional network of clinics conducted a rebranding: it changed its name, logo, and updated its website. Visually everything was perfect — a new corporate style, fresh design, and neat social media profiles. However, the JSON-LD Organization block still contained the sameAs field with old links to profiles that were still listed under the former name in three industry directories.
Result: Google Business Profile continued to display the old name in the knowledge panel, and when users asked ChatGPT "which clinic is better in the city," the model would name the company by its old name — because the training data and directories had not yet caught up with the rebranding. A routine visual audit would not have detected this issue: the website looked flawless. The problem only surfaced when checking the chain "markup → directory → AI response".
How to check whether brands are correctly recognized by ChatGPT and Gemini
Yes, this is part of a digital identity audit, not a separate discipline — it's just rarely included in classic checklists. The mechanism is simple: a set of 20 control prompts is compiled, arranged by stages of decision-making — from the general category request to comparison with specific competitors. Then two metrics are calculated.
- Share of Model — the share of prompts in which the model mentions the brand at all. A direct analog of Share of Voice from media planning.
- Citation Rate — how often the model confirms the answer by referencing the source. Here platforms differ significantly: Perplexity positions its citation accuracy at 97%, but independent audits show significant discrepancies between the source and the cited fact. The citation mechanism varies greatly between platforms: some embed links in most answers, others mention sources less frequently.
A low Citation Rate for ChatGPT does not mean the brand is not visible there — it means you need to directly verify the mention of the brand through prompts, rather than relying on transparent links. Tools like KlevelBrand formalize this into a Brand Score based on 10 parameters, evaluating consistency and visibility in a single figure from 0 to 100.
Minimal checklist for small business
A full three-layer audit requires resources that small businesses often lack. However, it is possible to conduct a quick diagnosis in one day if you limit yourself to priority points.
- Check the Organization block in JSON-LD: whether name, logo, and sameAs with valid links are present.
- Verify the name, address, and phone number on the website against the listings in Google and Yandex directories.
- Ensure that links from social networks lead to the relevant domain without redirects through old pages.
- Ask 5-10 direct questions about the brand in ChatGPT and Gemini and record whether the company is mentioned and what exactly is said.
- Check that most website visitors study it before purchasing and see the current design — data shows that 75% of people evaluate the reliability of a business based on the quality of the website design.
This set is sufficient to identify critical gaps between how the brand appears inside the company and what people and algorithms see from the outside. A full audit with Brand Score calculation and monitoring of Share of Model makes sense to add when the business expands beyond a single region or begins to seriously compete for visibility in AI search.
Recent articles
MegaFon, BMW, VTB and more 240+ companies. See how it looks in the showcase reel:
Our branding agency is approached to order:
Branding agency services
