Violet

Get named when your buyers ask the machine — cited in ChatGPT, Perplexity, and AI Overviews.

Entity clarity, extractable content, and earned third-party citations, run as one program on one weekly trend read. AI search is roughly 80% good SEO and earned media — we run the 20% that is genuinely different, and we tell you where the evidence stops.

What AI search actually involves.

  1. Entity and knowledge-graph clarityResolve your brand as one unambiguous entity — Organization schema, consistent naming, and a sameAs array — then earn a Wikidata entry, the highest-leverage move actually available. You cannot buy a Wikipedia page; notability needs independent coverage. Wikidata is the lever, and consolidation takes three to nine months.
  2. Content a model can liftLLMs chunk a page into passages and score each on its own. We write answer-first paragraphs of roughly 40 to 75 words under question headings, name entities instead of leaning on pronouns, and keep every section self-contained and fact-dense. This makes you extractable — it is a floor, not a lever that buys citations.
  3. Digital PR that earns citationsRoughly 84% of AI citations are earned media and about 0.3% are paid, so we build the coverage models trust: original-data studies journalists actually use, expert commentary, and thought leadership. This is where your real story, your SME, and your data are load-bearing — it cannot be manufactured.
  4. Presence on high-citation surfacesEach engine draws from different places — Perplexity leans on community, AI Overviews blend Wikipedia, YouTube, and Reddit, and only about 11% of cited URLs are shared across engines. We optimize per engine through genuine participation only. Reddit removes fake posts at scale and the FTC fake-review rule applies — we never plant anything.
  5. Crawlability hygiene for AIAI crawlers do not execute JavaScript — 500M-plus GPTBot fetches ran zero JS. A client-rendered SPA can rank in Google yet be invisible to ChatGPT, Claude, and Perplexity, so we spec SSR or pre-render to put content in the raw HTML, and we separate training bots from retrieval bots in robots.txt so you can opt out of training without deleting yourself from a citation engine.
  6. Citation monitoring across enginesA frozen buyer-intent prompt set — roughly 40 to 200 prompts across intent categories — run two to three times a week for four to six weeks per engine. We report ranges, not snapshots, because run-to-run drift is real. No tool exposes ground truth or guarantees improvement, and we say so.

How you get cited when someone asks AI

People discuss you across the web; the model reads those sources and names you in its answer. The work is making you the thing that gets read - and cited.

How the work goes.

  1. Be a corroborated entity, not a claim.

    Models cite what they can resolve and cross-check. We make you one unambiguous entity — consistent naming, a sameAs graph, and a Wikidata entry — because that is the highest-leverage move actually available. You cannot buy your way onto Wikipedia.

  2. Write passages a model can lift.

    LLMs score self-contained passages, not whole pages. We write answer-first, entity-named, fact-dense sections that a model can quote verbatim. This makes you extractable — it is table stakes, not a lever that buys a citation.

  3. Earn the mentions — never fake them.

    Roughly 84% of AI citations are earned media, so we earn coverage worth trusting through original data and expert commentary. We will not plant Reddit posts or broker reviews — the FTC fake-review rule and the platforms both catch it, and it is not who we are.

  4. Measure directionally — promise process, not placement.

    Measurement is immature and citations are volatile third-party model output. We promise the rigor of the process and honest reporting in ranges against your month-one baseline. We do not promise a specific citation, because no one can deliver one.

What physically lands in your inbox.

AI-visibility audit
Your share-of-model across every engine that matters — where you are cited, where you are mentioned but not cited, and where you are absent — plus the technical and entity gaps behind each, with an owner against every fix.
Frozen prompt library
The buyer-intent prompt set — roughly 40 to 200 prompts across intent categories — that we hold fixed run after run, so every measurement is comparable rather than a fresh snapshot.
Citation-gap analysis
Every target prompt sorted into three buckets — competitors cited where you are not, prompts where you are named but not cited, and prompts you already win — so priority is obvious and defensible.
Citation-target list
The earned-media plan: which publications, data assets, and community surfaces earn the third-party mentions models trust, mapped to the prompts each one is meant to move.
Monthly share-of-model report
Citation rate per engine, share of voice within your competitor set, and prompt stability — reported in ranges with the volatility caveats stated, all against your month-one baseline.

Why the answer names you - or does not

An AI answer is stitched together sentence by sentence, each phrase drawn from a source. The phrases it pulls from your pages are what put your name in the answer. Our job is making your pages the ones worth stitching from.

Questions buyers ask.

Is GEO just SEO with a new name?
Roughly 80% of it is good SEO — the technical eligibility, genuinely helpful content, and real authority that earn rankings also earn AI visibility. The GEO-specific 20% is real, though: passage-level extraction, entity and Wikidata work, per-engine surfaces, cross-engine monitoring, and fixing pages AI crawlers cannot render. It is also engine-dependent — AI Overviews pull about 76% of citations from the top ten, while roughly 80% of ChatGPT citations are not even in Google's top 100.
Can you guarantee we get cited?
No — and be wary of anyone who does. Citations are non-deterministic output from third-party models you do not control, and they are volatile: per-engine presence can swing hard within weeks, and roughly 70% of cited domains churn within six months. We guarantee the rigor of the process and honest reporting, not a placement.
How do you measure this?
A frozen prompt set — roughly 40 to 200 buyer-intent prompts — run two to three times a week per engine, reported as ranges rather than snapshots. Baselines run around 8 to 15% citation rate, optimized around 20 to 30%, leadership at 40%-plus. Run-to-run monthly noise sits around 40–60%, so we clear that noise floor before we claim a win. AI-referral traffic is the weakest signal — most of it is click-free and often lands as Direct — so we lean on citation rate, share of voice, and branded-search lift instead.
How long until it works?
Fast technical wins — robots.txt, SSR, entity schema — hit almost immediately, and new content can appear in one to two weeks. Entity consolidation takes three to nine months, and measurable share-of-model typically lands in the four-to-six-month window. It is never a straight line, and we never attach a fixed date to a citation.
Can we just pay to be placed?
Not for editorial citations — only about 0.3% of AI citations are paid, and the ad products some engines sell are separate, labeled, and not the same thing as being cited as a trusted source. We earn citations; we do not buy placement, and we would not trust an agency that claimed it could.
Why do you keep pointing us at Reddit and YouTube?
Because the engines cite them heavily — AI Overviews blend Reddit and YouTube, and Perplexity leans on community answers. But only through genuine participation: real answers from a real, disclosed account, and real video that earns its place. Reddit detects and removes planted posts at scale, and the FTC fake-review rule applies — so we never fabricate a presence.
Does this replace our SEO?
No — it extends it. AI search sits on top of the same technical eligibility, helpful content, and earned authority that SEO builds. Optimizing for AI features is, in Google's own words, "still SEO." We run them as one program rather than billing you twice for overlapping work.
Is an llms.txt file worth adding?
No. There is no evidence it moves citations — roughly 97% of llms.txt files receive zero requests, and Google's own guidance calls the idea speculative. Anyone selling it as a lever is selling you a file the engines ignore. The levers that matter are crawlability, entity clarity, extractable content, and earned mentions.

How the trend engine feeds AI search.

Every citation we chase is a question our engine saw your buyers start asking the machine before your category thought to answer it.

Where AI search fits the rest of the system.

SEO

The foundation AI search is built on — roughly 80% of getting cited is the technical eligibility and helpful content that also ranks. Same pages, same program.

Explore

Social media

The YouTube and community presence models cite gets built the same week as the feed — genuine participation, never planted posts.

Explore

Organic growth

The pillar AI search sits inside — search, social, and AI citations run as one workflow on one weekly trend read.

Explore

Want us to run your AI search?

30 minutes. We'll tell you where you're already cited, where you're invisible, and what it would honestly take to move your share-of-model.

Book a free strategy call