AI Brand Mentions: Two Columns, Not One Score
By William Zhu · Cofounder, InfiniSynapse · Last updated: 2026-09-10 · Last verified: 2026-09-10 · Methods: SEO Health
aivisibilitychecks at aimeetup.center — mention and citation as two columns, mention-only rows kept empty in the citation column. Not an official Google score. Observed CLI contract 2026-09-10 oninfinitegrowth@0.1.1:seo-health check --format jsoncan exit 0 whileissues[].statusiserror.
Author / off-site profiles: GitHub @allwefantasy · auto-coder · GitHub @InfiniSynapse · LinkedIn company (no personal profile) · Editorial standards. No personal LinkedIn or vendor badge. Product recognition: SEO Health Checker is one of two first-prize works in the InfiniSynapse × CSDN Vibe Coding contest (English recognition archive). InfiniSynapse co-hosted the contest. That list is not a review of this article.
Reviewed by: InfiniSynapse Data Team · method review 2026-09-10. First-party method review, not a third-party award.
Trust / COI: About · Corrections · Publishing principles · Privacy · Terms. SEO Health is commercial. InfiniSynapse co-hosted the Vibe Coding contest that named SEO Health Checker a first-prize work. The CLI contract below is observed. First-party desk tables stay illustrative. The InfiniSynapse Data Team publishes this desk method.
Table of Contents
- TL;DR
- What the two columns actually record
- A two-column framework for the log
- Two columns compared with a citation-rate API
- Landscape of mention and source slots
- How to log mention and citation
- Desk sample
- Scorecard
- Failure modes
- Cluster guides
- FAQ
- Conclusion
Cover: ./images/hero-ai-brand-mentions.png · Social: ./images/og-cover.png · Chart: ./images/chart-ai-brand-mentions.png
TL;DR
AI brand mentions are two columns: mention versus citation. Keep the empty citation cell. Do not invent a citation-rate API. Credits apply.
Direct answer: Treat AI brand mentions as two binary cells on one locked prompt: brand or URL present, source pointer present or empty. Store the empty citation cell. Do not recode that row as a percentage, a rank, or an official Google score.
Paste a public URL at aimeetup.center/seo-tools#check for the eight-module lights (title, meta, headings, density, images, links, tech, speed). Those lights are traffic colors, not a Google 100. seo-health aivisibility is a different command: it needs InfiniSynapse login and credits.
We evaluate AI brand mentions hands-on as the InfiniSynapse Data Team. Evidence pack or locked prompts go through aimeetup as the signed-in InfiniSynapse account. One-shot analysis, not training.
What the two columns actually record
Key definition: AI brand mentions are rows with a mention cell and a citation cell. A yes mention plus an empty citation is a real measurement. The empty cell is not a missing export, not a failed scrape, and not a license to invent a citation-rate API.
Observed CLI contract (2026-09-10, infinitegrowth@0.1.1): For this AI-visibility log pass, the same two public URLs were re-used so this page does not invent a private run. seo-health check https://infinisynapse.com/en/blog/data-agent-memory --format json --lang en exited 0 with page.title present and Image Alt Text = warning. The same command on https://example.com also exited 0 while issues[].status listed error on Meta Description, Robots.txt, and Sitemap.xml. Parse those status fields. Exit 0 is not a clean page. Desk tables below stay illustrative.
Independent citation: According to [SEO starter guide](https://developers.google.com/search/docs/fundamentals/seo-starter-guide), Google's SEO starter guide says the work is to help search engines understand and present your content. Google's SEO starter guide page is the third-party rule this write-up holds to. Illustrative desks below are not that rule.
People collapse those two cells because a fluent sentence feels like a win. The log has to refuse that collapse. A brand string in the answer is one event. A source pointer the engine shows as a reference is another event. AI brand mentions are the first event, with the second event on or off.
Wikipedia’s overview of citation as a scholarly practice still treats a citation as a pointer to a source, not as a synonym for being named. Honest AI brand mentions follow that split: name and source are not the same column.
Mention is a string you can defend
A mention is a string match you can show a reviewer: brand, product, or public URL appears in the answer text. Count aliases you froze with the prompt set. Do not invent a fuzzy “they meant us” rule after the run. AI brand mentions still need that defendable string, or the mention cell stays no.
If the page never states the entity in an extractable paragraph, the mention cell will stay quiet more often than you like. Write the closed claim first. Then log. The extractable-paragraph guide is the sibling for that input, not a substitute for the two-column sheet.
Citation is a source pointer, or it is empty
A citation is a source slot the surface displays. If the slot is missing, write empty. Do not paste a URL you wish had appeared. Crossref’s public registry exists because scholarly systems treat a citation as a resolvable record, not as a vibe. MDN’s blockquote element is the same split in HTML: a quote is not a cite attribute. AI brand mentions that fire without a resolvable slot stay mention-yes, citation-empty.
An empty citation cell can sit next to a perfect mention. That pair is the whole job. Recoding it as “partial credit” hides the only action you can take: earn a source slot, or accept that this surface named you from memory.
A two-column framework for the log
The working object is a locked prompt set plus two columns. AI brand mentions fail the moment someone deletes the empty citation cell to “clean the sheet.” You then lose the only week-over-week fact the desk can defend.
| Column | Pass value | Fail value | What you must not do |
|---|---|---|---|
| Mention | Brand, product, or URL in the answer | Absent or guessed | Fuzzy “they implied us” |
| Citation | Source pointer the surface showed | Empty | Invent a URL after the fact |
| Surface | One named row | Blended “AI percent” | Average four engines |
| Prompt | Locked set ID and locale | Weekly rewrite | Compare two questions |
Keep those four stacked. AI brand mentions earn their keep only when the citation column is allowed to stay empty.
The W3C annotation model still separates the body of a note from the target it points at. Replay that habit on the log: the answer text is one object, the source pointer is another. A body that names you and a target that never appears is still two cells.
Credits sit on the signed-in command
seo-health aivisibility, seo-health eeat, and seo-health gsc need InfiniSynapse login plus credits. seo-health check, seo-health audit, and seo-health links can run without that login. AI brand mentions sit on the credit side because they call a long task, not because Chrome failed to read title, meta, headings, density, images, links, tech, or speed.
Install the CLI with npm i -g infinitegrowth when you want the same command in a script. The bin is seo-health. The visibility command is seo-health aivisibility. It is not a free eight-light paste. Credits apply.
Two columns compared with a citation-rate API
A citation-rate API would claim a single percentage: citations divided by mentions, or citations divided by prompts. That number invents a third event. AI brand mentions refuse the division. The empty cell is the finding.
Classic blue-link rank asks where a URL sits on a results page. This log asks whether a locked prompt named you while the source list stayed blank. A position-3 blue link can coexist with an answer that names you and cites someone else.
Google’s SEO starter guide still explains crawlable pages, titles, and links. It does not ship a citation-rate API for answer engines. Stay on that honesty: eight lights and indexation are inputs. They are not a mention column and they are not a citation column.
If you already keep Perplexity as its own row, open the Perplexity mention-versus-citation lead for the surface rule. Stay here when the defect is the collapsed column, not the missing Perplexity row.
Landscape of mention and source slots
The landscape is a table of outcomes, not a vendor contest. AI brand mentions occupy one pair of cells: mention yes or no, citation present or empty. Mention plus citation occupies another outcome. Silence occupies a third. Gemini, Claude, and Google AI Mode get their own rows. Schema and sitemap samples are inputs, not ranks.
NISO’s information standards work still treats structured information as something you can exchange without rewriting the event. The two-column log is that habit: you exchange mention and citation as written, including empties.
A Gemini row can show mention plus citation while a second surface only names you. Do not average those outcomes. The Gemini surface guide keeps that engine on its own row. AI brand mentions stay the column rule that every row must obey.
OWASP’s note on content spoofing is the warning we point at when a deck fills the citation cell with a URL the engine never showed. Forging a source slot is not analysis. The two-column contract forbids that fill.
How to log mention and citation
This method H2 is the only place the product command is specified as a sequence. AI brand mentions are the two cells; seo-health aivisibility is the runner.
Submit the evidence pack or locked prompts at aimeetup while signed in. The run is one-shot analysis, not a training upload.
Step 1 — Freeze the two-column contract
Write six to twelve prompts you can stand behind in public. Freeze brand aliases you will count as a mention. Create two columns before the first run: mention and citation. AI brand mentions start after that freeze, not during a brainstorm that already rewrites the question.
Use a public URL. Do not paste secrets, tokens, or unpublished drafts into the checker.
Step 2 — Run seo-health aivisibility
Sign in. Spend credits. Run seo-health aivisibility against the locked set and the public URL. The two cells must appear as a row you can point at, not as a footnote you delete. If the product UI offers other surfaces, do not collapse them.
The eight-module paste at aimeetup.center/seo-tools#check can still run first without login. That pass does not write the two columns. AI brand mentions that pretend the free paste wrote citation cells have already failed the method.
Step 3 — Write empty citation cells in public
For each prompt, record mention yes/no and citation yes/no. When mention is yes and citation is no, leave the citation cell empty. A row that gets recoded as “0.5” has already failed. Store the raw answer snippet and the source list so a reviewer can dispute a false mention.
Do not call an empty cell a pipeline error. The empty cell is the finding. AI brand mentions that hide empties are a dashboard.
Step 4 — Re-run without collapsing the columns
Replay the same set on a later date. AI brand mentions earn their keep on the delta between two locked runs, not on a one-off screenshot. If you change a synonym, increment the set ID and start a new series.
Do not call this a rank recovery. There is no position to recover. There are two binary columns and a date.
Independent citation 2: According to [annotation model](https://www.w3.org/TR/annotation-model), this W3C document is an independent web-standards reference, not a ranking certificate. A second independent source, from W3C, keeps this page claim from resting only on first-party lights.
Desk sample: locked set versus ad-hoc chat
The table and chart below are an illustrative InfiniSynapse desk log, not a customer result and not an independent benchmark. Two dimensions: outcome (mention plus cite / mention only / silent) × prompt hygiene (locked set vs ad-hoc chat) on one public docs URL. Desk note ABM-705-20260909.
Illustrative two-dimension desk chart (outcome × locked set vs ad-hoc chat). Not a customer report. Social cut: ./images/og-cover.png.
| Outcome (illustrative) | Locked set | Ad-hoc chat |
|---|---|---|
| Mention + cite | 3 | 1 |
| Mention only | 4 | 2 |
| Silent | 1 | 5 |
Two dimensions: outcome × hygiene. The locked set produced four mention-only rows. That middle row is an empty citation cell, not a missing export. Ad-hoc chat produced more silence and fewer defendable empties, because the question drifted. Desk counts are illustrative. AI brand mentions that only publish the cited column would have hidden those four empties.
How to read the grouped bars
Each cluster is an outcome. Each bar is a hygiene state. Trust your live log if it differs. Do not treat these desk counts as a lift claim. AI brand mentions still need those two columns on your sheet.
Selection scorecard for two-column logs
Use this scorecard when someone demos a log that claims to handle AI brand mentions. Score the sheet, not the slide.
| Criterion | Pass | Fail |
|---|---|---|
| Columns | Mention and citation both stored | Single visibility number |
| Empty cells | Citation allowed to stay empty | Empties recoded as 0.5 |
| API claim | No citation-rate API sold | “Citation rate 62%” |
| Prompts | Locked set ID and locale | Weekly rewritten prompts |
| Credits | Login + credits disclosed | Implied free crawl of a surface |
| Google mix | No official Google score | Empty cell sold as a penalty |
A log that fails two or more rows is a dashboard, not a method. Eight lights do not fill a missing citation column. AI brand mentions still need the empty cell in public.
Failure modes that invent a citation rate
These failures show up in decks more often than in logs. AI brand mentions that survive them keep two columns and one empty cell.
Filling the citation cell after the run
A brand name in a sentence is not a source slot. Pasting your homepage into the citation cell because “they clearly used us” is spoofing. Leave the cell empty and write the extractable paragraph again.
Averaging empties into a rate. Dividing citations by mentions invents a third event. A week with four mention-only rows and two cited rows is not “33% citation rate.” It is four empties and two source slots. AI brand mentions refuse that division.
Using a Google ranker as a substitute. Buying a classic rank tracker and labeling the export as a mention log is the fastest way to lie to yourself. AI brand mentions are not that export. Keep blue-link software for blue links.
Cluster guides for this visibility expansion
This page is the two-column cluster in the existing expansion. The table names sibling jobs. The links sit in surrounding sentences.
| Hub slug | Job on this cluster |
|---|---|
| extractable-paragraph-seo | Closed claim before you log |
| perplexity-search-rank-tracking-tools | Perplexity as its own row |
| gemini-brand-visibility | Gemini on its own row |
AI brand mentions sit beside those jobs. They do not replace a surface row and they do not replace eight lights.
Keep the empty citation cell in the weekly log
Paste a public URL at the live checker, then run the locked prompt set and store mention and citation as two columns.
Run SEO Health CheckerFrequently Asked Questions
Is an empty citation column a measurement error?
Bottom line: No. An empty citation cell next to a yes mention is still a row of AI brand mentions. Store the empty. Do not recode it as a pipeline failure or a 0.5.
Do I need credits to store both columns?
Bottom line: Yes for seo-health aivisibility. That command needs InfiniSynapse login and credits. The eight-module paste at aimeetup.center can run without login and does not write AI brand mentions.
Can I invent a single citation-rate percentage?
Bottom line: No. A citation-rate API is not a product we ship and not a number this desk will invent. AI brand mentions stay two binary columns.
Does a brand name in a sentence count as a source slot?
Bottom line: No. A mention is text. A citation is a source pointer. AI brand mentions that never show a pointer stay mention-yes, citation-empty.
Conclusion
AI brand mentions are a two-column log: mention yes or no, citation present or empty. Freeze the prompt set, run seo-health aivisibility with credits, and keep the empty cell. Do not invent a citation-rate API. Do not sell an official Google score.
If you need the signed-in account that spends those credits, open the InfiniSynapse web app once, then return to aimeetup for the checker. Re-run the same set. Keep AI brand mentions as cells you can defend, not as a percentage you wish you had.