When someone asks ChatGPT for a product in your category, it doesn’t always crawl websites in real-time.
Its first move is to pull from what it already knows about you and your competitors from its existing knowledge.
Clear and recognizable entities in AI training data are just as important as having the most authoritative and optimized website.
This shift means your webpage might rank #1 in classic search, but if your brand isn’t well-structured for entities, AI might overlook you entirely in the answer.
The rules we’ve relied on for decades don’t fully apply when machines create answers. They draw on their own knowledge and real-time data from sites, including yours.
You’re about to learn what this means, why it matters, and what you can do about it.
What Are Entities in AI Search?
An entity is a “thing” that search engines and AI models can recognize, understand, and connect to other things.
Think of entities as the building blocks that AI uses to construct answers. In other words, gigantic relational databases.
Let’s use email marketing company Omnisend as an example.
Through the lens of a database, Omnisend isn’t just a website with pages about email marketing. It’s a network of connected entities:
Use cases: “welcome series,” “abandoned cart recovery”
Here’s what the entities look (hypothetically ) like to a large language model (LLM):
These records become the foundation for AI answers.
LLMs do more than just find keywords on your page. They also retrieve entities, place them in vector space, and choose the ones that best answer your question.
Vector space explained: It’s a mathematical method that AI models use to understand relationships between concepts. Imagine a 3D map where similar items group together. For example, “Apple,” the company, is close to “iPhone” and “Tim Cook.” Meanwhile, “apple,” the fruit, is near “banana” and “orchard.”
For example, ask Google: “What’s the best email marketing tool for my Shopify store?”
You’ll see brand entities like Klaviyo, Omnisend, Brevo, Mailchimp, Privy, and MailerLite mentioned. This makes sense because the entities are closely related in the AI’s understanding.
Notice: the brand mentions aren’t linked to the websites. It’s just building the answer and then linking to the brand SERP on Google.
Why Entities Matter More Than Websites
AI models are constantly mapping relationships between entities when serving up answers.
When someone types “best email marketing tool for Shopify,” LLMs spread out the query. They turn that one question into multiple related searches.
Think of AI doing lots of Google searches at the same time.
The system simultaneously explores “What integrates with Shopify?”, “Which tools handle abandoned carts?” and “What do ecommerce stores actually use?”
Your brand can appear through any of these paths, even if you didn’t optimize for the original query.
Classic SEO relied a lot on keyword density and page authority.
But AI uses dense retrieval, where it’s looking for semantic meaning across the web, not just word matches on your page.
Dense retrieval explained: AI systems focus on meaning, not just exact keywords. They find related content, even if different words are used.
A Reddit comment that clearly explains “We switched from Klaviyo to Omnisend because the Shopify integration actually works” carries more signal (assuming the model prioritizes authentic discussions) than a page stuffed with “best email marketing Shopify” keywords.
The AI understands the relationship between the entities (Klaviyo, Omnisend, Shopify) and the context (switching, integration quality).
PR folks have been fighting for this moment: mentions without links still count.
For the longest time, we’ve obsessed over backlinks as the currency of SEO.
But AI systems recognize when brands get mentioned alongside relevant topics, using these as relationship signals.
So when Patagonia appears in climate articles without a hyperlink, when Notion shows up in productivity discussions on Reddit, when your brand gets name-dropped in a podcast transcript — these all strengthen your entity in AI’s understanding.
Here’s a real example that clarified this for me:
Microsoft OneNote often shows up high in AI recommendations for “note-taking tools.”
In ChatGPT:
In Perplexity:
And in Google AI Overviews:
But EverNote dominates Google’s number one ranking spot for “note taking tools”.
Why?
OneNote’s integration with the Microsoft ecosystem means it gets mentioned constantly in productivity discussions, enterprise software comparisons, and Office tutorials. This creates dense entity relationships in AI training data.
Evernote, by contrast, has focused on SEO and earned strong backlinks that dominate traditional search rankings.
How Entities Get Recognized
So how does Google (and other AI systems) actually know that Omnisend is an email marketing platform and not, say, a meditation app?
The answer sits at the intersection of structured data, human conversation, and pattern recognition…at massive scale.
Entity Databases and Product Catalogs
Google maintains what they call Knowledge Graphs and Shopping Graphs.
Other AI systems have similar entity databases, just with different names.
The idea is the same: huge databases that map every product, company, and person along with their attributes and relationships.
When Nike releases the Pegasus 41, it doesn’t just become a new product page on Nike.com. It becomes an entity in Google’s Shopping Graph, connected to “running shoes,” “Nike,” “marathon training,” and hundreds of other nodes.
The system knows it’s a shoe before anyone optimizes a single keyword.
Human Conversation as Training Data
AI systems learn just as much from informal mentions as they do from structured markup.
When an Outdoor Gear Lab review casually mentions testing Patagonia’s Torrentshell 3L against the expensive Arc’teryx Beta SL, that relationship gets encoded.
When a podcast guest says, “I moved from Asana to Notion for task and project management,” this competitive link adds to the training data.
Reddit and Quora have become unexpectedly powerful for entity recognition. (Google explicitly stated they’re prioritizing “authentic discussion forums” in their ranking systems.)
A single comment on why someone picked Obsidian over Notion for knowledge management matters more than you might realize.
These platforms capture what websites struggle to do: real people sharing real decisions with real context.
Multimodal Recognition
AI systems extract entities from audio and video. They do this by turning speech into text through transcription.
Every mention in a transcript, every product on screen, and every comparison in a talking-head segment is processed.
A 10-minute YouTube review of project management tools turns into structured data that compares ClickUp, Notion, and Asana. It includes feature comparisons and maps out use cases.
The New SEO Power Dynamic
You can’t game entity recognition the way you could game PageRank.
You can’t manufacture authentic Reddit discussions. You can’t fake your way into natural podcast mentions. The system rewards genuine presence in genuine conversations, not optimized anchor text.
Think about what this means:
Your engineering team’s conference talk that mentions your product’s architecture? That’s entity building.
Your customer’s YouTube walkthrough of their workflow? Entity building.
That heated Hacker News thread where someone defends your approach to data privacy? Entity building.
We’ve spent the longest time optimizing for robots. Now the robots are optimized to recognize authentic human discussion. (Ironic.)
5 Ways to Optimize Your Brand for Entities (Not Just a Website)
Using Omnisend as an example, here are five approaches for evaluating and optimizing entity presence in AI-powered search results.
1. Assess Your Entity Foundation
To start, you need a baseline understanding of your current entity relationships.
For Omnisend, this means mapping how AI systems currently categorize them relative to competitors.
Begin by verifying schema markup across key pages.
Testing Omnisend’s homepage with the Schema Markup Validator shows they use Organization and VideoObject schema.
And the Organization schema is relatively basic.
Omnisends competitor, Klaviyo, uses Organization schema as a container for multiple software offerings.
Klaviyo’s approach maintains brand-level authority while declaring specific software categories and capabilities. This potentially gives them stronger entity associations for queries about email marketing, SMS marketing, and marketing automation.
Next, check your entity presence in major knowledge sources like Wikidata and Crunchbase.
On Wikidata, Omnisend’s records are OKAY.
There’s basic info, like what Omnisend does, the industry, inception date, URL, and social media profiles.
But Klaviyo, again, is all over it. They have multiple properties for industry, entity type, URLs, offerings, and even partnerships.
There’s a clear opportunity for Omnisend to update its Wikidata with more details.
2. Test Query Decomposition
AI systems break down queries into entities and relationships. Then, they may try multiple retrievals.
For example, in Google Chrome, I prompted ChatGPT:
“What’s the best email marketing tool for ecommerce in 2025? My priority is deliverability.”
In the chat URL, copy the alphanumeric sequence after the /c/ directory. For me, it was 68d4e99e-4818-8332-adbd-efab286f4007.
Note: You need to be logged into ChatGPT to get this sequence
Right-click on the page and click “Inspect”.
Choose the “Network” tab, paste the alphanumeric sequence in the filter field, and reload the page.
In the “Find” section, search for “search_model_queries“. Then, click on the search results.
Each decomposed query represents a different competitive pathway.
Omnisend might surface through deliverability discussions, but miss general tool comparisons.
Mailchimp could dominate broad searches while competitors own specialized angles.
This explains why you appear in AI answers for searches you never optimized for. The semantic understanding creates visibility through unexpected entity relationships rather than keyword matching.
You can check this yourself. Run the extracted queries in separate chats and note which brands appear where.
But maybe don’t build a strategy around exploiting this technique.
The methodology depends on undocumented functionality that OpenAI could change without notice.
Important finding: Simple queries produce simple results. When I prompted “Best email marketing tool for ecommerce,” it triggered exactly one internal search with basically the same language. No decomposition.
3. Map Competitive Entity Relationships
Traditional SEO competitive analysis asks “Who ranks for our keywords?”
Entity analysis asks “When do AI systems group us together?”
I tested this with Omnisend to understand when they appear alongside different competitors.
I ran 15 variations of email marketing queries through Google AI Mode to see which brands consistently appear together.
Note: I tested logged out, using a VPN set to San Francisco, in private browsing mode to minimize personalization bias.
I began with simple terms like “best email marketing for ecommerce” and “abandoned cart recovery tools.” Then, I tried different angles like “email automation for Shopify stores.”
Here’s what I found:
Query Context
Omnisend Present
Most Co-Mentioned
Klaviyo Present
Ecommerce email
5/5 queries
Klaviyo, Mailchimp
4/5 queries
General email
5/5 queries
Mailchimp, Brevo
2/5 queries
Deliverability focus
2/5 queries
Brevo, Mailchimp
0/5 queries
Omnisend appeared in 12 of 15 total queries — stronger entity presence than I expected.
But mentions shifted dramatically by context.
In ecommerce discussions, Klaviyo dominated as the top tool.
In general email marketing, Mailchimp took over as the main reference point.
The mention order revealed something important. Klaviyo appeared first in 5 of 5 ecommerce queries, with more positive language around their positioning.
Omnisend routinely ranked second or third. This suggests they’re part of the discussion but not at the forefront.
Here’s what’s interesting:
Klaviyo completely disappeared from deliverability-focused queries while Omnisend maintained some presence.
This shows entity relationships are radically contextual.
Being the leader in ecommerce email doesn’t mean presence in deliverability conversations.
4. Optimize For Entities in Your Content
Entity recognition works best when it has context-rich passages. This helps AI systems extract and understand information more easily.
Take generic descriptions like “Our automation features help ecommerce businesses increase revenue through targeted campaigns.”
An AI system may struggle to identify which product you mean, its automation features, or how it compares to others.
Compare that to: “Omnisend’s SMS automation integrates with Shopify’s abandoned cart data to trigger personalized recovery messages within 2 hours of cart abandonment, without requiring manual workflow setup.”
This version establishes multiple entity relationships (Omnisend → SMS automation → Shopify integration → abandoned cart recovery) within a single extractable passage.
LLMs prefer to use their training data for answers. But when they pull info from the web, strong entity connections help a lot.
You’re reducing friction for both bots and human readers.
As a test, run key passages from your most important pages through Google’s Natural Language API to see what entities get recognized. This can also be video scripts.
Content with strong entity density tends to get cited more often than content requiring additional context.
5. Build Strategic Co-Citations
Entity authority builds through consistent mention alongside relevant entities in trusted sources. This moves the focus from link building to building relationships where natural comparisons happen.
For Omnisend, this means being present in authentic discussions. It’s about genuine comparisons, not forced mentions, that strengthen specific relationships.
A Reddit thread comparing “Klaviyo vs Omnisend for Shopify stores” carries a different entity weight than appearing in generic “email marketing tools” content.
The specific context (Shopify integration) strengthens both brands’ association with ecommerce email marketing.
The most valuable co-citations happen in:
Reddit discussions comparing tools for specific use cases
YouTube reviews demonstrating multiple platforms
Industry roundups grouping tools by specialization
Podcast discussions of marketing technology stacks
This Reddit thread shows strategic co-citation in action. The original post creates dense entity relationships (Klaviyo → Omnisend → pricing → Shopify store). While the comment adds even more context (pricing concerns → business scaling → “pretty good” user experience).
The discussion goes way beyond optimized content. It’s genuine decision-making that strengthens both brands’ entity associations with ecommerce email marketing.
This approach emphasizes genuine participation. Your category is discussed and evaluated by actual users who make real decisions. This is better than having artificial mentions in content made mainly for search engines.
Moving Forward with Entity SEO
If you’ve built a strong brand across various channels, you’ve laid the foundation.
Quality SEO is still crucial.
Genuine mentions in industry talks, real customer chats, and multi-channel distribution matter too.
Begin with your key product line. Organize it well, track its appearances in AI responses, and then expand to other entities.
Backlinko is owned by Semrush. We’re still obsessed with bringing you world-class SEO insights, backed by hands-on experience. Unless otherwise noted, this content was written by either an employee or paid contractor of Semrush Inc.