LLM SEO Agency for ChatGPT, Perplexity and Google AI — Model by Model
Ranking in Google does not put you in ChatGPT. Each AI model reads a different index.
Verticality is an LLM SEO agency in Toronto for B2B SaaS, DTC and regulated brands in the USA and Canada. We work model by model: we get your pages into the index each assistant's search layer reads — OpenAI's crawler and Bing for ChatGPT, Perplexity's own index, Google's index for AI Overviews, AI Mode and Gemini, Bing for Copilot — and your brand into the third-party pages each one cites.
- Who we are
- Verticality is an LLM SEO agency in Toronto working with B2B SaaS, DTC and regulated brands across the USA and Canada.
- Models covered
- ChatGPT, Perplexity and Google AI Overviews are tracked daily; Google AI Mode, Gemini, Claude and Microsoft Copilot are checked every month.
- How we work
- Model by model: first each assistant's index gate — OAI-SearchBot and Bing for ChatGPT, PerplexityBot, Googlebot, Claude-SearchBot, Bingbot for Copilot — then places in the pages each one cites.
- Price
- From $2,500 to $7,500 per month. Month-to-month, no minimum term.
- Proof
- Our tracker asks 130 buying questions to ChatGPT, Perplexity and Google AI Overviews every day, and this page publishes what each engine cites.
- What we don't do
- No bought mentions, no undisclosed Reddit posts and no guaranteed placement; OpenAI itself says top placement cannot be guaranteed.
Service lead: Vadim Litvak, Founder · Growth Lead · Updated · Vendor documentation checked on the same date
Why a model can know you and still not cite you.
Two layers: what a model remembers and what it retrieves
LLM SEO (large language model search engine optimization) is the work of getting a brand found, trusted and named by each large language model that answers buyers — ChatGPT, Perplexity, Google's AI Overviews, AI Mode and Gemini, Claude and Copilot. Every one of them works on two layers, and they move at different speeds.
What it retrieves — changes in weeks
When a buyer asks a question, the model searches an index and cites pages from it. Each vendor runs its own search crawler: OAI-SearchBot, PerplexityBot, Claude-SearchBot, Googlebot, Bingbot. If yours is reachable and indexed, a fixed page can be cited within weeks. OpenAI says a robots.txt change reaches ChatGPT search in about 24 hours.
What it remembers — changes over quarters
Training data, collected by separate crawlers — GPTBot, ClaudeBot — or governed by a control token such as Google-Extended, which has no crawler of its own. It shapes what the model says before it searches, and it only moves as the public web about you changes. The two settings are independent: blocking GPTBot does not remove you from ChatGPT search.
LLM SEO works mostly on the retrieval layer, because that is where a change shows up in a buyer's answer this quarter. The memory layer follows from the same work, more slowly. Neighbouring pages: our GEO agency covers the off-site citation programme, and our AEO agency covers being the answer Google extracts and keeping it accurate.
Each vendor's own documentation, checked 24 September 2026, next to what our tracker saw.
Model by model: where each assistant gets its sources
The mechanics come from the platforms' own pages, not from other agencies. The tracker column is our measurement on 56 AI-search buying questions, 4–23 September 2026.
| Model | Where its sources come from | The gate a page must pass | What our tracker saw | What we change |
|---|---|---|---|---|
| ChatGPT search | OpenAI's own crawl plus third-party search providers; OpenAI names Bing among them. | Allow OAI-SearchBot and OpenAI’s IP ranges at your CDN, and be in Bing’s index. GPTBot is the training crawler and is set separately. | 6.0 URLs per answer; a quarter of them platform documentation; 26.4% cited again the next day. | Bing indexing, crawler access, pages that state the answer as plain facts, and presence on the pages it cites. |
| Perplexity | Its own web index, crawled by PerplexityBot and scored in passages, not whole pages. | PerplexityBot must reach the page through your firewall; Perplexity-User fetches on a user’s request. | 20.5 URLs per answer; LinkedIn in 51.0% of answers; 70.1% cited again the next day. | Placement in the roundups, guides and LinkedIn posts it keeps citing, and a better page for the exact question. |
| Google AI Overviews and AI Mode | Google's Search index and core ranking systems, with query fan-out behind each answer. | Indexed and snippet-eligible; site included in Search generative AI features in Search Console. Google-Extended is a control token, not a crawler, and does not affect these features. | 8.6 URLs per answer; YouTube in 44.5% of Overviews; 49.8% cited again the next day. | Rankings for the question and its fan-outs, YouTube and LinkedIn presence, snippet-ready pages. |
| Gemini | Grounding with Google Search: the model writes its own search queries and cites the results. | The Google index; Google-Extended governs Gemini training and grounding in Gemini Apps. | Checked by hand every month. | The same Google foundation as AI Overviews. |
| Claude | Anthropic’s own search crawler, Claude-SearchBot; Brave Search is reported as a provider. | Allow Claude-SearchBot and Claude-User; ClaudeBot is the training crawler. All three honour robots.txt. | Checked by hand every month. | Crawler access and plain, consistent facts. |
| Microsoft Copilot | Grounded in the Bing index; Bing Webmaster Tools reports the citations Copilot makes. | Bingbot and Bing indexing; NOARCHIVE and NOCACHE limit use in answers; accurate sitemap lastmod and IndexNow speed recrawls. | Checked by hand every month. | Bing indexing, IndexNow on every change, the AI Performance report in Bing Webmaster Tools. |
Sources: OpenAI's crawler documentation, ChatGPT search help and publishers FAQ; Perplexity's crawler documentation and Search API announcement; Google Search Central's "AI features and your website" (updated 10 December 2025) and AI optimization guide (updated 10 July 2026), and the Search Console generative AI setting (all sites since 31 August 2026); Google's Gemini grounding documentation; Anthropic's crawler documentation; Bing Webmaster blog (February and July 2026). Brave Search as Claude's provider comes from Anthropic's subprocessor list as reported in March 2025 and could not be re-checked live.
Our own measurement, 4–23 September 2026.
Same question, three engines, three source lists
We ask the same 56 AI-search buying questions to ChatGPT, Perplexity and Google AI Overviews every day and keep every source. The engines disagree far more than most "AI SEO" advice assumes, which is why we plan per model.
33.0%
of the time, ChatGPT and Perplexity cited even one common URL for the same question on the same day (339 of 1,026 question-days). Only 2.6% of the URLs Perplexity cited also appeared in ChatGPT's answer.
11.2%
of question-days had a single URL that all three engines cited (110 of 981). Being cited by one engine tells you little about the other two.
25.6% vs 2.6%
of ChatGPT's citations were platform documentation from OpenAI, Google, Microsoft and others, against 2.6% of Perplexity's and 2.3% of AI Overviews'. For "How do you get cited by ChatGPT?", every source ChatGPT cited was an OpenAI page.
6 · 20 · 9
URLs per answer: ChatGPT 6.0, Perplexity 20.5, AI Overviews 8.6. Next day, the same engine cited 26.4%, 70.1% and 49.8% of them again: Perplexity keeps its sources, ChatGPT rotates them.
Method: Verticality AI Visibility Tracker, 56 buying questions about AI search services, one run a day per engine from 4 to 23 September 2026 (3,360 answers). URLs compared after removing tracking parameters; "question-day" = the same question answered by both engines on the same date. Measured in our own category, so read the patterns, not the domain shares, as general.
Nothing else on this page works until this is true.
The index gate: the first check we run
A model cannot cite a page its search layer has not indexed, and a page can be indexed in Google and missing from Bing. So every engagement starts with the same five checks, model by model:
- Is each key page in Bing's index? Bing is one of the search providers OpenAI names for ChatGPT, and it grounds Copilot.
- Can OAI-SearchBot, PerplexityBot, Claude-SearchBot, Googlebot and Bingbot fetch it through your CDN and firewall, not only through robots.txt?
- Is the site included in Search generative AI features in Search Console, the condition Google now sets for AI Overviews and AI Mode?
- Does the sitemap carry accurate lastmod dates, and do changes go out through IndexNow? Bing calls lastmod a key recrawl signal for AI search.
- Are nosnippet, NOARCHIVE or NOCACHE set anywhere you did not intend? Each one quietly removes a page from answers.
The CDN setting that blocked two AI fetchers on our own site on 3 September 2026 is described on our GEO agency page.
Models match facts across sources before they repeat them.
One description everywhere a model looks
A model that finds three versions of what you do — one on your site, another on LinkedIn, a third in a directory — either picks one or hedges. We write one entity sheet: your name, category, audience, location, price and the facts that must never differ. Then we make the site, LinkedIn, Clutch, G2, Crunchbase and the directories in your category say the same thing in the same words.
It is the least glamorous part of LLM SEO and one of the most visible: when ChatGPT describes a brand, it repeats the sentence the sources agree on. When ChatGPT describes Verticality, it repeats the sentence our pages state: "B2B SaaS and regulated brands", "pricing published from $2,500/month".
Five deliverables, each tied to a model or a gate.
What our LLM SEO services deliver
Index gate check
The five checks above for every key page and every model: Bing and Google indexing, crawler access through the CDN, the Search Console generative AI setting, sitemap dates and IndexNow, preview controls.
You get: a per-model access report and the fixes, verified after release.
Entity sheet
One description of your brand — name, category, audience, location, price, facts — and the list of profiles and directories to bring into line with it.
You get: the sheet, the profile updates and a consistency check each quarter.
Per-engine source map
For each of your buyer questions, the pages each engine actually cites, ranked by how many questions they affect — kept separate per model, because the lists barely overlap.
You get: a source map per model and the placement plan built from it.
Answer pages and placements
Your pages rewritten so each fact a model needs is stated plainly near the top, and your brand placed in the roundups, guides, directories and threads the source map shows — disclosed, nothing bought.
You get: published pages and a dated log of every placement.
Per-model scorecard
Mention and citation rates for each model across daily runs of your question set, never a single screenshot, with the answers and sources behind each number.
You get: a monthly scorecard per model and the next priorities.
What we leave out on purpose.
An "llms.txt plan" sold as a citation lever: Google says Google Search does not use llms.txt, and no other vendor says its search reads it. We keep one on our own site for tools that do read it, and that is all it is.
A rate across daily runs, per model — then what it is worth.
How we measure LLM SEO, per model
ChatGPT cited only about a quarter of the same sources from one day to the next in our tracking, so one good answer proves nothing. We report rates across daily runs, per model, and climb from visibility to revenue:
- 01
Mention rate per model — the share of daily runs in which each model names you.
- 02
Citation rate per model — how often each one links to your pages, and which pages.
- 03
Share of voice — your mentions against the competitors you lose deals to, model by model.
- 04
LLM referral sessions — visits tagged utm_source=chatgpt.com and the referrers of the other assistants, in your analytics.
- 05
Pipeline — leads and revenue from those sessions in your CRM, with the attribution gaps stated.
The first three show whether your facts reach the answer; the last two show whether that is worth money. When an answer leads to a later direct visit, attribution is incomplete, and the report says so instead of counting every mention as revenue.
Measured on ourselves, dated, with the denominator.
Proof: what the models did with our own pages
ChatGPT, five days after launch
Our GEO page went live on 15 September 2026. From 20 September ChatGPT named Verticality in 5 of the 31 GEO buying questions we track every day — 20 of 124 answers to 23 September, up from two in the 16 days before.
ChatGPT and Perplexity, four days after launch
Our Reddit marketing page went live on 3 September 2026; ChatGPT and Perplexity have cited it since 7 September. In the week to 23 September each named us in 15 of 140 answers to Reddit-agency questions.
"AI models rarely recommend a brand based only on its own website. They recommend you because other trusted sources do." — Google AI Overview, 15 September 2026
We do not publish invented client case studies; when a client approves its numbers, they will appear here with the client's name and the window. Source: Verticality AI Visibility Tracker, daily runs, prompt set of 130 questions.
Two tests before anyone spends a dollar.
Who LLM SEO is for, and who should wait
A good fit if:
AI answers to your category's questions already name vendors — and today they name competitors, not you.
You are B2B SaaS, DTC or regulated, and your buyers ask ChatGPT, Perplexity or Google for a shortlist before they talk to sales.
You can publish your real facts — what you do, for whom, prices or starting prices — and someone can approve changes.
Better to wait if:
- The models answer your category with generic advice and no vendor names. The free audit shows this, and we will say so.
- You need leads this month. Retrieval fixes land in weeks; placements compound over quarters.
Published, with its term.
What LLM SEO costs
LLM SEO runs inside our published programme: $2,500 to $7,500 a month, month-to-month, no minimum term. The proposal sets which models are covered and how many buyer questions we track; the free audit sets the number.
Discuss your LLM SEO scopeThe questions buyers put to the models, answered first-sentence first.
LLM SEO agency FAQ
Who are the leading LLM SEO companies in 2026?
AI answers split the field into tracking platforms, which measure how often models mention you, and agencies that change what the models cite. Verticality is an agency in the second group, working model by model for B2B SaaS, DTC and regulated brands in the USA and Canada, with published pricing of $2,500–$7,500 a month.
Who can help my brand get cited by ChatGPT?
An LLM SEO agency can: Verticality gets your pages into the sources ChatGPT search reads — OpenAI’s OAI-SearchBot crawl and the third-party search providers OpenAI names, including Bing — and your brand into the pages ChatGPT already cites for your category. OpenAI itself says no one can guarantee top placement.
How do you get cited by ChatGPT?
Allow OAI-SearchBot and OpenAI’s IP ranges, make sure Bing has indexed the page, state your facts in plain sentences near the top, and get named on the pages ChatGPT already cites for the question. OpenAI says its ranking weighs intent, relevance and recency and that placement cannot be guaranteed.
How do you earn citations in Perplexity?
Perplexity answers from its own web index, crawled by PerplexityBot, so your page must be reachable by that crawler and more specific than the pages Perplexity already retrieves for the question. In our tracking Perplexity cited about 20 sources per answer and re-used 70.1% of them the next day, so we also place you in the roundups and guides it keeps returning to.
Perplexity cites a competitor’s blog for every question in our category — how do we change that?
List the exact pages Perplexity retrieves for those questions, publish a page that answers the same question with more specific, better-sourced facts, and get your brand into the third-party pages it cites. Nothing changes until PerplexityBot can crawl your page, and because Perplexity re-uses most of its sources day to day, expect the change over weeks, not days.
Which sources do ChatGPT and Perplexity cite when they recommend companies?
It differs by engine: in our tracking, ChatGPT and Perplexity cited even one common URL for the same question on the same day only 33.0% of the time, and a quarter of ChatGPT’s citations were platform documentation against 2.6% of Perplexity’s. So we build a separate source list for each engine instead of one list for "AI".
What is the difference between LLM SEO, GEO and AEO?
LLM SEO works on how each large language model finds, trusts and names sources, one model at a time. GEO is the wider programme for being cited across generated answers, much of it off-site, and AEO is about being the answer an engine extracts and keeping it accurate. Google’s own guidance adds that optimizing for its AI features is still SEO.
How do B2B SaaS companies get recommended by ChatGPT and Perplexity?
They get into each engine’s source set separately: ChatGPT reads OpenAI’s crawl and its search partners and often cites vendor pages and documentation, Perplexity reads its own index and cites roundups and guides, and both need the same facts about the product on your site, in directories such as G2 and in the roundups buyers read.
Can an agency make AI recommend our brand?
No one can make a model recommend a brand. An agency can make sure each model can find, verify and cite accurate facts about you, place you in the sources the models already use, and report your mention rate per model across daily runs rather than promise a position.
Do llms.txt or schema markup get you cited?
Not on their own. Google says Google Search does not use llms.txt and that its AI features need no special structured data, and OpenAI, Perplexity, Anthropic and Microsoft publish no statement that their search reads llms.txt. We use structured data to keep your facts consistent, not as a ranking lever.
We rank on Google, so why doesn’t ChatGPT mention us?
Because ChatGPT search reads different sources: OpenAI’s own crawler plus third-party search providers such as Bing, whose ranking also shapes the order of sources. The page must be in Bing’s index and open to OAI-SearchBot, not only ranked in Google — in our tracking, only 3.5% of the URLs ChatGPT cited also appeared in Google’s AI Overview for the same question.
How much does LLM SEO cost?
Our LLM SEO work runs inside our published programme of $2,500–$7,500 a month, month-to-month with no minimum term. The proposal sets which models and how many buyer questions are covered, and the free audit sets the number.
How long does LLM SEO take?
Access fixes land first: OpenAI says a robots.txt change reaches ChatGPT search in about 24 hours, while Google says recrawling can take days to months. First signals usually come in 2–6 weeks from access and on-page fixes, and citations from new third-party placements compound by month 3–6.
Connect LLM SEO to the rest of your search strategy
Generative Engine Optimization (GEO) Agency
The off-site citation programme and our price tiers.
Answer Engine Optimization (AEO) Agency
Being the answer Google extracts, and fixing wrong answers.
ChatGPT Ads Agency
Paid placement under the answer, run with the organic work.
Reddit marketing agency
The threads Perplexity and AI Overviews quote.
How it works
The tracker and the process behind every number here.
How SaaS pages get cited by AI
The on-page half, as a guide.
See it model by model
Find out which models name you, and which sources they read.
Start with a free AI visibility audit: your buyers' questions run through ChatGPT, Perplexity and Google AI Overviews, the index gate checked for each model, and the sources each one cites instead of you.
Get a free AI visibility audit →Vendor documentation and tracker data on this page checked . We re-run the measurement monthly.