AI search keywords in your Aiprobase profile are not SEO keywords. The logic is different, and conflating the two leads to keywords that don't work.
Here's the distinction: SEO keywords are optimised for how people type into a search bar. AI search keywords reflect how an AI agent interprets a user's request and looks for a match. The user might say "find me someone who can help automate my invoicing process" — the AI agent translates that into a concept, not a keyword string. Your keywords need to map to those concepts.
This section is available on the Verified tier and maps directly to knowsAbout in Schema.org — the field that tells AI agents what you have genuine expertise in.
What AI search keywords actually do
When an AI agent receives a query, it identifies the underlying need and looks for profiles where knowsAbout contains matching concepts. Your keywords are not matched character-by-character like a search engine — they're matched semantically.
This means "invoice automation" and "accounts payable automation" can both match a query about "automating my billing workflow". Specificity matters more than volume. Three precise keywords outperform twenty generic ones.
How to think about your keywords
Start from the query, not from your own vocabulary.
Ask yourself: what would a client say to an AI agent when they need what I offer? They won't say "B2B SaaS marketing consultancy services". They'll say:
- "Find someone who can help me get more demos for my SaaS product"
- "Find a consultant who's worked with fintech startups"
- "Find someone who knows Klaviyo and email marketing for e-commerce"
The keywords that match those queries are: SaaS demand generation, fintech startup marketing, Klaviyo email marketing, e-commerce retention.
Write down 10–15 queries your ideal client might say. Extract the concepts. Those are your keywords.
What makes a good keyword
Specific beats generic. "Marketing" is too broad to be useful. "B2B content marketing for SaaS" is matchable. "Email automation for Shopify stores" is matchable.
Tool and platform names are valuable. If you work with specific software — Salesforce, Figma, AWS, Webflow, QuickBooks — include them. Clients often search by tool: "find a Webflow developer", "find someone who knows Salesforce CPQ".
Industry terms work better than job titles. "UX design" is a job title. "Mobile app onboarding design" is a problem. "Conversion rate optimisation for checkout flows" is a problem. AI agents are better at matching problems than job titles.
Process and methodology terms are underused. "Agile project management", "design sprints", "jobs-to-be-done research", "double-entry bookkeeping" — these are precise signals that match specific queries.
What to avoid
Avoid adjectives. "Expert", "experienced", "professional", "innovative" — these carry no matchable information. Every profile claims them. They add noise.
Avoid your own brand name. Your brand name is already in your profile. It doesn't need to be repeated in keywords.
Avoid generic category labels. If you're a graphic designer, "graphic design" as a standalone keyword is too broad to differentiate you. Add specificity: "brand identity design", "print design for packaging", "editorial illustration".
A practical example
A freelance data analyst working primarily with e-commerce companies:
Too generic:Data analysis, Excel, SQL, reporting, dashboards, business intelligence
Better:E-commerce sales analytics, Shopify data analysis, customer cohort analysis, retention reporting, SQL dashboards, Google Looker Studio, abandoned cart analysis, LTV modelling
The second set matches specific queries: "find someone who can analyse my Shopify data", "find a freelancer who can build a retention dashboard", "find someone who knows cohort analysis for e-commerce".
How many keywords to add
The profile accepts 3–20 keywords. Aim for 8–12. Below 8, you're leaving coverage gaps. Above 15, you risk diluting precision with filler terms.
Review your keywords when your work changes — new specialisations, new tools, new industries you've entered. Your profile updates immediately on save.
Keywords and your descriptions work together
Keywords amplify your descriptions, they don't replace them. An AI agent reads your description first for context, then uses your keywords as additional matching signals. A keyword that also appears naturally in your full overview carries more weight than a keyword that appears nowhere else in your profile.
For how to write descriptions that work alongside your keywords, see How to write your business description for AI search.
FAQ
Should my keywords match what I'd use for Google SEO?
Not necessarily. Some overlap is fine, but the logic is different. SEO keywords are shaped by search volume. AI search keywords should be shaped by the specific queries your ideal clients would make. Optimise for precision, not volume.
Can I use multi-word phrases as a single keyword?
Yes, and you should. "Mobile app UX design" is a better keyword than "UX" alone. Phrases that describe a specific capability or specialisation are more useful than single words.
How quickly do keyword changes take effect?
Immediately. Your JSON-LD updates on save and your profile is re-indexed shortly after.
