What does Evertune do?
Evertune is the first Generative Engine Optimization (GEO) platform built to explore, measure, act and advertise across the entire AI customer journey, connecting brands directly to ChatGPT Ads and programmatic advertising partners. Where most GEO tools sample each prompt once a day, Evertune samples every prompt 100 times per model across 11+ AI models, delivering statistically significant visibility data at half the cost of competing platforms.
Evertune's agents do the work for you: a prompt agent mines over 150 million real user conversations to find the exact questions buyers ask, an insights agent tells you what to do next each week, and an ads agent builds a complete ChatGPT campaign around your visibility gaps. From there, Evertune closes the loop with website optimization, data-driven content creation, source-level influence mapping and paid activation through affiliate and programmatic AI retargeting partners. Founded by the team that pioneered programmatic advertising at The Trade Desk, now building the next marketing channel.
What is Generative Engine Optimization (GEO)?
GEO, or generative engine optimization, is the practice of increasing how often and how favorably a brand appears in answers generated by AI models such as ChatGPT, Gemini, Claude and Perplexity.
Where SEO works to rank a page in a list of links, GEO works to make the brand part of the answer itself. Three things have to go right: the brand gets mentioned when a buyer asks an open question with no brand name in it, the brand is described accurately, and the pages that support that description are ones the model actually reads. GEO operates on three separate levers, and they behave differently:
- Base model knowledge is what a model absorbed during training. It moves on training cycles, so months.
- Live retrieval is what a model finds when it searches the web mid-answer. It moves in weeks.
- Paid placement fills the gap while organic presence builds.
Why does Evertune sample each prompt 100 times per model?
Because a single response is one draw from a distribution, not a measurement.
LLMs are probabilistic, so the same prompt returns different answers. Most GEO platforms sample each prompt once per model per day, which cannot distinguish a pattern from an anomaly. Evertune asks each prompt 100 times per model instead, which brings the margin of error to roughly one point at the overall level and two points at the topic level.
Clustering semantically related prompts into topics adds a second layer of stability by smoothing prompt-level variation, which is what makes topic comparisons reliable over time.
Can Evertune show me which sources AI cites in my category?
Yes. Content Analytics reports every domain and URL AI models return as a source, then separates the ones that matter from the ones that merely appear often.
Frequency alone is misleading, so each source is scored on Topic Relevance (how much it shapes AI's understanding of your category) and Brand Relevance (how much it discusses your brand, weighted by sentiment). Those two scores sort every source into three groups:
- Owned URLs: your own properties showing up as sources, the easiest to optimize.
- Strength URLs: high influence and they mention you favorably
- Opportunity URLs: high influence and your brand is absent. The highest-ROI targets for PR and content placement.
Partner Connect then shows which affiliate platforms can reach those domains, so the list becomes a campaign rather than a spreadsheet.
Can Evertune tell me what to fix on my own website?
Yes, through two features that answer different questions.
Site Audit evaluates how well AI crawlers can reach and parse your pages, scoring bot permissions, sitemap quality, page load performance, JSON-LD, title tags, meta descriptions and heading hierarchy at both page and domain level. It crawls up to 1,000 pages and can be scoped to a subfolder or subdomain.
Bot Analytics shows which AI bots actually visited, which pages they crawled, how often they returned, and what response codes they received. It separates training bots from search and user bots, which tells you whether you are feeding base model knowledge, live retrieval, or neither. Blocked requests, high error rates and a low percentage of the site crawled are the three findings worth acting on immediately.
Can you advertise on ChatGPT through Evertune?
Yes, you can run paid advertising on ChatGPT through Evertune. Our Ads Agent builds and manages those campaigns from your AI visibility data, targets the conversations where your brand isn't yet recommended, and reports in more detail than the native ChatGPT interface.
Can Evertune create content for me?
Yes. Our Content Agent generates messaging, keywords and ready-to-use blog copy built around the gaps in your data.
It works backwards from your scores. Pick the Consumer Preference where your AI Brand Score is lowest, and Content Studio returns three strategies for that topic, each with a key message and suggested keywords, drawn from roughly 50 candidate messages tested against the models. From there our Content Agent generates a full draft in one of six templates at lengths from 250 to 3,000 words.
Add your brand guidelines once and everything generated follows them. Supply up to 10 URLs of your own content to teach it your tone of voice. Drafts are editable in-platform and export as Markdown or HTML.
Can Evertune track how AI recommends my products?
Yes. Shopping Intelligence tracks the product-level experience, not just the brand mention.
ChatGPT now surfaces product cards with images, prices, reviews and purchase links in response to questions like "what's the best espresso machine under $200." Shopping Intelligence reports when your products appear, how visible they are against competitors, which retailers are featured alongside them, and what prices the models are quoting.
Three uses that come up most: finding SKUs with low or zero Product Visibility Score, which usually signals thin descriptions or missing attributes rather than a demand problem; fact-checking the prices AI is quoting against your actual prices, since a stale retailer promotion becomes the number in the shopping card; and auditing whether your key distribution partners are showing up at all.
Which LLMs and AI search engines does Evertune monitor?
Evertune currently monitors ChatGPT, ChatGPT Search, Gemini, Gemini Search, Google AI Mode, Google AI Overviews, Meta AI, Claude, CoPilot, Perplexity and DeepSeek. We regularly add new models.
Where do Evertune's prompts come from?
Every prompt is grounded in EverPanel, Evertune's proprietary consumer intelligence panel of more than 150 million real user conversations, demographically weighted to reflect the composition of the internet.
That matters because over 80% of AI prompts are phrased uniquely and have never been asked the same way twice, so keyword lists do not describe real behavior. EverPanel shows the language people actually use, the intent patterns behind it, and how often each topic comes up.
Product Category Analysis trackers apply three validated prompt types: Word Association (aided, brand name in the prompt), AI Brand Index (unaided, engineered like a survey measuring unaided awareness) and Consumer Preferences (attribute-by-attribute comparisons). Suggested Prompts builds the bank dynamically from EverPanel, clustering semantically related prompts into real intent groups for your category.
What reports does Evertune provide?
Evertune provides insights into prompt behavior and AI model usage (Prompt Volumes, Prompt Research, AI Usage); brand monitoring (AI Brand Index, Word Association, Sentiment Analysis, Consumer Preferences, Source Insights, Shopping Intelligence); and website optimization (Site Audit and Bot Analytics).
What is the difference between base model and consumer app data?
Base model data is what a model knows from training. Consumer app data is what it says after running a live web search. Evertune is the only platform that reports both.
Base model responses are reachable only through direct API integration with the model provider, which is the only way to isolate foundational knowledge. Consumer app responses come from the interface a buyer actually uses, where retrieval is in play.
A brand that performs well in the base model but poorly in the consumer app has an SEO and retrieval problem. A brand that is weak in both has a brand-building problem. Base model knowledge also matters more over time, because agent experiences are built on top of it and inherit its brand preferences.
How do AI models decide which brands to recommend?
Models recommend the brands they have seen described most consistently, by the most credible sources, in the context of the question being asked. An LLM is not looking up a ranking. It is predicting the most probable answer based on patterns it has absorbed, which means visibility is a function of how your brand appears across the whole web rather than how one page is optimized.
Why do AI models give different answers to the same questions?
LLMs are probabilistic, meaning they generate each response token by token instead of retrieving a stored one, so the same prompt returns different answers.
Four things introduce variation: probabilistic sampling during generation, conversation history and memory, changes in the live search index between queries, and model updates over time.
The practical implication is a measurement problem. A single spot check is one draw from a distribution, not a reading. A brand that appears in half of responses will look present or absent depending on which day you check. Sampling the same prompt repeatedly is what separates a pattern from an anomaly: Evertune samples each prompt 100 times per model, which brings the margin of error to roughly one point at the overall level and two points at the topic level.
Do AI responses come from training data for live web search?
Both! When a user asks an AI model a question, the model determines if it knows enough from its base knowledge to generate a response. It evaluates whether it has sufficient information from its training data alone. If it can't find all the information it needs, the AI runs multiple queries in a search index, visits the URLs, gathers more information to supplement its existing knowledge, and then generates a response based on both its training AND the retrieved information, citing selected sources it visited. This process is called Retrieval Augmented Generation (or RAG).
This is why only tracking consumer app insights like most GEO platforms do tells an incomplete picture. You need to understand what the model inherently knows about your brand through its training data AND what it learns about your brand in real time.
What type of content do AI models cite most often?
Listicles, by a wide margin. In an analysis of the top 25,000 cited URLs, listicles accounted for half of heavily cited URLs and 63% of citations by volume.
That analysis covered citations across ChatGPT, Copilot, Gemini, Google AI Mode, Google AI Overviews and Perplexity (AI search loves listicles: what 25,000 URLs reveal about citations, updated May 2026). Ranked lists dominate the format, making up 71% to 86% of listicles depending on the model. Unranked lists are a distant second.
Five other formats earn citations reliably: feature articles, primers that define a concept, short blogs, case studies with quantified outcomes, and how-to guides.
Commercial content is not disqualified. Sponsored and affiliate pages get cited when they carry genuine editorial value and disclose the relationship. One exception: do not publish a listicle that ranks your own brand against competitors. Google has signaled intent to penalize the format and FTC rules prohibit presenting controlled content as independent review.
Why would different versions of the same model produce different results?
Different versions of the same model produce different results because newer versions are usually bigger, trained on more recent information, and built with improved techniques. Even when they seem similar, each version has learned different patterns from different data, so they naturally respond differently to the same prompts.
How do you get a brand recommended in ChatGPT?
1. Confirm AI crawlers can reach your site and are not being blocked or served errors.
2. Identify the domains and URLs AI models cite for your category, then flag the high-influence ones that never mention you.
3. Earn presence on those pages through PR outreach, review sites, affiliate partnerships and community engagement.
4. Publish owned content that answers real category questions directly, in the formats models extract from.
5. Repeat the same core claims in varied phrasing across multiple credible sources rather than duplicating one page.
Content that lives only on your own domain carries less weight than the same claims appearing across industry publications, review sites and partner blogs. You cannot spam your way in. Models weight meaning, credibility and consistency, not keyword frequency.
What metrics are important to measuring AI visibility performance?
- Visibility Score (0 to 100): the percentage of responses that mention your brand
- AI Brand Score (0 to 100): mentions weighted by position. Position one carries 100% weight, position two 90%, position three 81%, position four 73%, each step down 10% of the one before it. A score of 100 means first position in every response.
- Average Position: where your brand typically lands in the list.
- Share of Answer: your proportion of all brand mentions, summing to 100% across every brand.
Why did my brand stop appearing in ChatGPT recommendations?
A model update is the most likely cause, and it can shorten the recommendation list across an entire category without anything changing on your site.
Evertune ran about 220,000 prompts across nine product categories to compare GPT-5 mini with GPT-5.4 mini. Answering from foundational knowledge with no search, the newer model recommended 37% fewer brands on average. Every category narrowed, from a 16.7% drop in basketball shoes to 59.2% in CRM. Airlines fell from 59 brands to 34, CRM from 49 to 20, handbags from 88 to 59 (ChatGPT Gets Pickier, March 2026).
As models get more selective, the shortlist gets shorter and mid-tier brands fall off first. Three checks before treating it as your problem: did competitors drop at the same time, did the change appear in the base model or only in the consumer app, and did it coincide with a published model version change?
Which AI models should brands track?
Track the models your buyers use. The set that covers most consumer and B2B behavior today include ChatGPT, Gemini, Google AI Mode, Google AI Overviews, Copilot, Perplexity, Claude, Meta AI and DeepSeek.
What's the difference between Google AI Overviews and Google AI Mode?
AI Overviews are automatic summaries at the top of a normal Google results page. AI Mode is a separate conversational search experience the user chooses to enter.
AI Overviews launched broadly in the US in May 2024. They appear without any opt-in on queries where Google decides a summary adds value, sit above the blue links with citation cards, and skew toward quick informational intent. Reach is the defining feature: every Google user sees them on eligible searches.
AI Mode began rolling out in 2025 as a distinct interface, closer in use to ChatGPT or Perplexity than to a results page. It handles multi-step research, follow-up questions and synthesis across angles, which is where high-consideration buyers do their comparison work.
Neither is a standalone model. Both combine Google Search infrastructure with Gemini.
