AI site search pricing gets confusing fast because vendors aren’t charging for the same thing.
Algolia meters search requests and records. Google’s Agent Search bills by query and indexed data, with extra charges for more advanced generative answers. Cludo goes in another direction with annual pricing and separate AI layers.
That matters more than the headline plan price. A university where people type full questions into the search box behaves very differently from a developer portal where users jump between product names, error messages, and documentation pages. A government website with a lot of short navigational searches is different again.
So before comparing prices, you need to know what people are actually doing on your site.
How We Picked These Tools
This list is aimed at teams choosing AI search for a public-facing website.
Every product here does more than rank matching pages. Each offers some form of generated answer based on indexed content, though the way those answers are delivered, managed, and priced varies quite a bit.
We compared pricing, implementation, answer citations, conversational search, and the kinds of websites each product appears to suit. We haven’t tried to turn every vendor into one made-up monthly cost. The billing units are too different, and your own search behavior will have more influence on the final bill than a generic traffic estimate.
The Six Tools at a Glance
| Tool | Best fit | What the bill counts | Published entry price | AI answers |
| AddSearch | Organizations that want keyword search, AI answers, and conversational search in one website platform | Plan tier and selected capabilities | $119/month for Keyword Search, $700/month for AI Answers, billed annually | Source-cited AI Answers plus context-aware AI Conversations |
| Algolia | Developer-led teams that want extensive control over search infrastructure | Search requests and records | Free up to 10,000 requests/month, then $0.50 per additional 1,000 on Grow | Agent Studio for newer AI integrations, NeuralSearch on Elevate |
| Agent Search by Google Cloud | Teams already comfortable building inside Google Cloud | Queries and indexed data | Enterprise Search $4 per 1,000 queries | Core Generative Answers included, Advanced Generative Answers cost another $4 per 1,000 |
| Cludo | Universities, government organizations, and other large public websites | Flat annual core fee plus AI layers | Quoted, with AI layers from $8,000/year each | AI Summary, AI Chat, and AI Mode |
| Coveo | Large enterprises with complex content and search permissions | Website search capacity and generative query entitlements | Custom enterprise pricing | Relevance Generative Answering and Conversational Search |
| Kapa.ai | Documentation and technical support environments | Answers and retrieval queries | 14-day free trial, then quoted | Cited answers across documentation and support interfaces |
1. AddSearch
AddSearch is the closest fit here for the public website itself. Under the AddSearch Experience name, Keyword Search, AI Answers, and AI Conversations sit inside one search and content discovery platform.
In practice, those searches don’t all look the same. Someone might want a specific form and expect a normal results page with the right document near the top. The next visitor types out an entire question and gets a short answer instead. If they ask a follow-up, AI Conversations keeps the context rather than making them start again.
The control over source material is useful on large sites where old information tends to hang around. The website owner decides which parts of the site are included in search and answers. AI responses link back to their sources, so a visitor can open the material behind the answer rather than taking it on trust.
The logs are probably the less obvious part. A month of unanswered questions can tell you quite a lot about the website itself. Maybe visitors keep looking for a policy that exists but is buried four levels deep. Maybe everyone asks the same question because the relevant page explains it badly. Search data starts becoming a content audit without anyone calling it one.
Keyword Search starts at $119 per month when billed annually. AI Answers starts at $700 per month on annual billing. AI Conversations and the complete AddSearch Experience use custom pricing.
You also don’t have to rip out an existing search provider before testing the AI layer. AddSearch positions AI Answers and AI Conversations as capabilities that can work alongside existing search infrastructure, which is useful if a team wants to add an AI answer engine without rebuilding its whole search setup first.
Most complete implementations take around two to four weeks, depending on the site and level of customization. AddSearch is SOC 2 Type II certified and says customer content isn’t used to train AI models.
2. Algolia
Algolia makes more sense once you stop thinking of it as a finished search product and start thinking of it as infrastructure your team builds around.
The free plan includes 10,000 search requests and 50,000 records each month. Grow also includes 10,000 requests, then charges $0.50 per additional 1,000. It includes 100,000 records, with another $0.40 per 1,000 beyond that.
Grow Plus raises additional search requests to $1.75 per 1,000 and adds AI Ranking, AI Synonyms, personalization, and other AI-assisted features. NeuralSearch is part of the annual Elevate plan.
Algolia’s generative layer is changing. Ask AI still appears in its search products, particularly around DocSearch, though new AI integrations are being moved onto Agent Studio rather than the older standalone Ask AI API.
That distinction matters if you’re comparing what a new deployment looks like today rather than reading an older feature page.
The bigger consideration is still implementation. Developers generally own the index structure, interface, relevance setup, and integrations. Algolia gives them a lot of control, which is useful when search is part of the product itself.
For a marketing or content team hoping to manage most of the search experience without ongoing engineering help, that freedom can turn into extra work.
3. Agent Search by Google Cloud
Google’s product is now called Agent Search, following the earlier Vertex AI Search branding.
Under Google’s general pricing model, Enterprise Search costs $4 per 1,000 queries and includes Core Generative Answers. Advanced Generative Answers add another $4 per 1,000 user-input queries.
The first 10,000 queries each month are free, apart from the advanced generative add-on.
Indexed data is billed separately once you move beyond the free storage allowance. Google currently provides the first 10 GiB of Agent Search index storage at no charge, then applies storage pricing above that.
The pricing is unusually transparent for an enterprise search product. The setup is less plug-and-play.
Agent Search sits inside Google Cloud. Someone still needs to configure the data store, connect the content, create the search experience, and manage the surrounding cloud setup.
If your technical team already runs plenty of workloads in Google Cloud, that probably isn’t much of a hurdle. A university communications team or association marketing department without cloud engineers is buying something quite different from a managed site search platform.
4. Cludo
Cludo comes up often in public-sector and university search because its commercial model is built more like an annual software contract than a metered search service.
Its core search pricing is a flat annual rate based on the organization and its needs rather than website traffic or raw search volume.
The AI features sit on top of that. AI Summary, AI Chat, and AI Mode each start at $8,000 per year, per AI layer. That kind of pricing is easy to dismiss if you’re only lining up entry plans in a spreadsheet. It looks different when a university budget has to be approved for the year and nobody wants to explain an unexpected usage spike to finance halfway through it.
Web teams can manage search behavior through Cludo’s own interface, while the platform also supports federated indexing and AI layers grounded in site content.
Cludo offers US or EU hosting for its AI layers too. That may matter more to a European public organization than shaving a small amount off the query price.
Pricing beyond the published AI-layer starting point requires a quote.
5. Coveo
Coveo sits at the heavier enterprise end of this list.
Its website product is designed around large search environments where content may be spread across several systems and access rules matter. Coveo’s own product entitlements describe website plans in units of 100,000 queries per month rather than a small self-serve monthly package.
Its Relevance Generative Answering layer produces answers from securely indexed enterprise content and adds citations back to the material used to produce the response.
There is also a conversational layer. Coveo’s Conversational Search keeps the context of follow-up questions and runs each turn through the same retrieval and permissions framework.
That becomes useful in a large organization where “search the website” isn’t really one simple index. You may have information across a CMS, support content, a portal, and other repositories, with different people allowed to see different things. Coveo is built for that level of complexity.
The trade-off is that this isn’t a lightweight website-search purchase. Pricing is custom, its website licensing is based around enterprise query capacity, and the generative AI products are additional extensions to the main Coveo subscription.
AddSearch itself positions Coveo as the stronger fit when a large enterprise has content spread across multiple systems and has the search team and budget for a more involved deployment. That’s a fair distinction here too.
6. Kapa.ai
Kapa.ai has a much narrower focus than the rest of the list.
Its natural home is technical documentation and support, where someone might paste an error message, ask how an API behaves, or try to work out why a particular configuration isn’t doing what the docs said it would.
The platform indexes sources such as documentation, websites, GitHub, forums, and YouTube. Growth plans also support private or authenticated sources including Confluence, Notion, Zendesk, Salesforce, and Google Drive. The answers come with citations to the underlying content.
Its free trial lasts 14 days and includes one index, 5,000 crawled pages, 5,000 retrieval queries, and 1,000 answers. Growth and Enterprise plans are sized to the customer and quoted.
Growth also adds features such as continuous syncing, zero data retention, and PII masking. Enterprise goes further with options including SSO, SCIM, RBAC, and regional hosting.
For a general public website, much of that specialization may go unused. On a documentation portal where people regularly arrive with very specific technical questions, it makes a lot more sense.
Which Pricing Model Fits Your Website?
Before you compare plans, look at a month of your own search data.
The first useful split is between short keyword searches and full questions.
“Financial aid” is a search.
“How do I apply for financial aid if I’m an international student?” is closer to an answer-engine query.
Those two people may be looking for similar information, but the products above won’t necessarily meter or handle those interactions in the same way.
Then look beyond query volume.
A cheap API can become expensive if developers have to spend weeks building the interface, maintaining the index, and tuning everything around it. A higher software fee sometimes looks more reasonable once the vendor handles more of the setup and your web team can manage routine changes itself.
Content complexity matters as well. A public website with a few thousand pages is a different job from an enterprise search environment pulling information from several systems with different access permissions.
Thirty days of search logs is usually enough to get started. Separate short searches from full questions, look at how often people reformulate queries, and note where the content itself lives.
Then compare the products against that reality.
It’s less tidy than picking the cheapest number in the table.
It’s much more useful.
FAQ
What Is AI Site Search?
AI site search combines website search with generated answers based on indexed content.
Depending on the platform, a visitor might get a normal results page, a direct answer with links to its source material, or a conversational experience where follow-up questions keep the previous context.
What Is an AI Answer Engine for Websites?
An AI answer engine responds directly to a natural-language question using information from the website or other approved content sources.
The stronger implementations also show citations, which lets the visitor check the source material behind the answer. Some products combine that with keyword and conversational search inside the same broader website experience.
How Much Does AI Site Search Cost?
There isn’t one useful average because the vendors charge in very different ways.
On this list, published entry points range from Algolia’s free tier and Google’s monthly free query allowance to $700 per month for AddSearch AI Answers and $8,000 per year for an individual Cludo AI layer. Coveo and Kapa use quote-based pricing for the relevant enterprise plans.
Those numbers still don’t tell you the full cost. Query volume, generated answers, indexed content, implementation work, and the way visitors actually search can all change the final figure.
