The Google PSE Shutdown: 6 Site Search Alternatives for Higher Ed

Ten years ago, Google told higher education to go find a new search engine. It is doing it again.

In February 2016, Google announced it was discontinuing the Google Search Appliance. No new hardware sales after 2016, support winding down through 2018, and complete discontinuation by 2019. Hundreds of institutions that had built campus search around a rack-mounted box in a data center got about two years to figure out what came next. Google offered no migration path and no replacement product.

A lot of campuses landed on Google Custom Search Engine, because it was free, and it was Google. The University of Utah published what that actually cost them: 130 separate Custom Search Engines configured, more than 190 search boxes migrated across campus, and a project that ran until the summer of 2018.

Custom Search Engine is now Programmable Search Engine. It’s sunsetting too.

January 1, 2027. That is when the Custom Search JSON API stops serving traffic. Google closed it to new customers in 2025 and gave everyone still on it a hard date to move. No extension has been announced.

Google's own documentation points existing customers toward Google's paid solution, Vertex AI Search, which it describes as a good fit for searching up to 50 domains.

Fifty domains is a reasonable ceiling for a mid-size company. Your university passed it sometime around 2012.

The 2016 lesson is worth saying plainly. Campuses that took the free Google product are now doing the migration a second time. Campuses that bought a platform built for higher education are not.

This article is about the search box on your own website, the one prospective students use to find a program page or a financial aid deadline. 

What actually changed

Google splits Programmable Search Engine into four offerings. The differences matter more than most teams realize, because the free tier is the one with ads on it.

	Standard Element	Non-profit Element	Paid Element	Custom Search JSON API
Cost	Free	Free	$5 per 1,000 queries	$5 per 1,000 queries
Ads	Yes	No	No	No
Google branding	Optional	Required	Optional	None
Daily cap	None	None	None	10,000 queries
Availability	Everyone	Non-profits only	Everyone	Existing customers only

Source: Programmable Search Engine Overview, Google for Developers.

The free Standard element serves third-party ads on your own results page. The free Non-profit element strips the ads but requires Google branding and is limited to non-profits. Going ad-free any other way means paying, at $5 per 1,000 queries.

Plenty of public institutions have been running third-party ads on their .edu search results for years without anyone flagging it. If that describes your campus, the sunset is doing you a favor by forcing the conversation.

The JSON API, the programmatic route most developers used, is the piece with the deadline. Per Google's own documentation, it provides 100 free queries per day, with additional requests at $5 per 1,000 and a hard ceiling of 10,000 queries per day. That pricing now applies only to existing customers until the service is discontinued on January 1, 2027. Google's note on the same page directs those customers to Vertex AI Search as a favorable alternative for searching up to 50 domains, and routes anyone who needs full web search to a contact form.

This is the second endpoint to go. The Custom Search Site Restricted JSON API, the version built for engines searching 10 or fewer sites with no daily query limit, stopped serving traffic on January 8, 2025. Google pointed those customers to Vertex AI Search then too.

There is a third limit that rarely surfaces until it bites. Google's Help Center advises keeping a Programmable Search Engine to no more than 20 unique sites, with every domain you add counting toward that total, and warns that engines exceeding 20 sites may return fewer results than a site: search on Google.com would.

Twenty sites on the product you have today. Fifty on the product Google wants you to move to.

A university is not one website

Digital sprawl makes site search infrastructure, not a feature.

The 2026 benchmark from our Digital Sprawl Report puts the average university main site at 5,000 to 8,000 pages, before counting departments, labs, and centers. Those pages are not interchangeable. A search box on a .edu has to handle degree programs, individual courses, faculty profiles, course catalogs buried in PDFs, events, news, and policy documents. A prospective student, a current student, and a registrar all type "nursing" and all want something different.

Sprawl is how research universities are built, and no deadline is going to clean it up. IT is decentralized by design. Faculty own their own sites. Grant-funded microsites outlive the grants that paid for them. Athletics runs on a separate stack. The medical center runs on another one entirely. Every one of those properties has a legitimate owner who is not handing over the keys because a search platform asked nicely.

Most site search tools are built for one website. A university runs hundreds of them.

How to evaluate a replacement

Six things separate the tools that survive contact with a campus from the ones you replace again in three years.

1. Who controls relevance. Can a communications person pin a result, add a synonym, and fix "financial aid" on a Tuesday afternoon? Or does it become a dev ticket that lands in a sprint three weeks out? Search relevance in higher ed changes with the academic calendar. If tuning requires engineering time, it does not happen.

2. AI built into the search experience. Does AI actually improve discovery, or is it just a chatbot bolted onto the site? Look for natural-language understanding, intent-based search, smarter relevance, and AI-generated answers or summaries grounded in the institution’s own content. AI should work alongside traditional search controls, analytics, synonyms, and relevance tuning so the institution stays in control.

3. Higher ed content types. Are programs, courses, faculty, events, and PDF catalogs indexed as distinct types with their own fields and filters, or are they all just pages? A university is not a collection of web pages. Its search platform should understand the difference.

4. Domain and subdomain scale. Can it crawl hundreds of properties across a decentralized institution without a hard cap or a cost cliff that kicks in at domain 51? What works for one website may fall apart at university scale.

5. Search analytics as content strategy. Query data tells you what prospective students tried to find and could not. That is a content gap report, a UX research tool, and a direct signal of student intent. It should be usable by marketing and content teams, not buried in an IT dashboard.

6. Accessibility and procurement. VPAT, WCAG conformance, SOC 2, and a security review your CIO will clear. None of the capabilities above matter if the platform cannot make it through institutional procurement.

Six alternatives to Google PSE

	Higher ed focus	Handles 100s of domains	Marketer-friendly tuning	Gen AI summaries	Search-focused company	Setup effort
SearchStax	Yes	Yes	Yes	Yes	Yes	Low to moderate
Cludo	Yes	Yes	Yes	Yes, sold as separate AI layers	Yes	Low to moderate
AddSearch	No	Yes	Yes	Yes	Yes	Low, fastest to launch
Site Search 360	General purpose	Yes	Partial	Limited	Yes	Low
Funnelback (Squiz)	Yes	Yes	Partial	Yes	No	Moderate to high
Algolia	No, ecommerce led	Yes	No, API-first	Yes	Yes	High
Vertex AI Search (baseline)	None	No	No	Yes	Yes	High

1. SearchStax

SearchStax’s largest vertical is higher education, with customers that include the University of Arizona, The Ohio State University, and Texas Christian University. partnerships with major higher education technology providers such as Acquia and Terminalfour further reinforce that focus. That specialization also shows up in the product itself, with unified search designed to bring together pages, courses, documents, people, events, and news as distinct content types rather than treating everything as a single flat index.

Marketers can manage synonyms, pin results, and adjust ranking directly. SearchStax also conforms to WCAG requirements, an important consideration for higher education institutions where accessibility is both a user experience and compliance requirement.   The analytics carry the other half of the argument. Query data shows what students searched for and failed to find, which turns your search box into a content gap report. For institutions moving off PSE under deadline pressure, it is the closest thing on this list to a purpose-built replacement.

2. Cludo

The no-code Experience Builder puts tuning in the hands of web and marketing teams without developer dependency, and the platform is accessible, which clears a real hurdle in public institution procurement.

The AI capabilities are packaged as separate layers on top of search, so the quote you get for keyword search is not the quote for AI answers. The search products are also quote-only, which lengthens an already long procurement cycle. If your requirement is a single index spanning hundreds of properties with a managed Solr layer underneath, compare the architecture carefully rather than the feature list.

3. AddSearch

Ad-free and on-brand by default, with AI answers drawn from your own indexed content and source citations on every response. It will tell a visitor it cannot find something rather than inventing an answer, which is not a given in this category.

A fast option to get live but also general purpose, so you build the higher ed content types yourself instead of inheriting them.

4. Site Search 360

Site Search 360 built its business on the previous Google search retirement, positioning itself as the ad-free replacement when Google Site Search was discontinued. That history makes it a natural landing spot, and it drops in with a line of JavaScript.

Lighter on AI features and on the procurement documentation your security review will ask for. Good fit for a single site, less so for a federated campus.

5. Funnelback (Squiz)

Squiz is a DXP provider that includes a site search option as part of their SaaS offering. Squiz is the one company on this list that is not completely focused on site search. With a deep customer base in Australia and the UK and a presence in North America. It understands institutional content and handles scale.

The lift is heavier, and the conversation usually arrives attached to a broader digital experience platform discussion. If you are only replacing a search box, that can be more procurement than you signed up for.

6. Algolia

Algolia is largely e-commerce focused. It is a solid experience for developers with good documentation and speed. If you have engineers who want to own the search layer, this is the one they will pick.

That is also the catch. Tuning happens through the API, which puts every relevance change back in the development queue.

And the baseline: Google Vertex AI Search

Vertex is Google's own recommended path, so it belongs in the comparison even though it is not a fit for most campuses. Google's Custom Search JSON API documentation describes it as a favorable alternative for searching up to 50 domains. It also requires full Google Cloud setup, OAuth configuration, and enterprise-tier commitment.

For a single-site institution with tight domain governance, it is worth pricing out. For a research university with 1,500 digital properties, it’s probably not feasible.

Your migration timeline

Start with an inventory of your domains and subdomains. Most teams cannot name the real number, and finding it out is half the project. From there, shortlist and demo on your own content, not a sandbox. Then budget 90 days minimum for procurement and security review, longer if your system office touches the contract. Implementation and reindexing take another 60 days on top of that, and reindexing is not a switch. It is a crawl, a review of what came back wrong, and a second crawl.

That is five months of work against a hard cutoff, and the calendar does not negotiate. Every week spent deciding whether this is urgent comes out of the reindexing window at the far end, which is the one phase you cannot compress. Miss it and you are not migrating, you are recovering, during the two months of the year when prospective students are trying to find your application deadline. The inventory of your domains and subdomains  is the only step you can take this week without a budget line or a vendor call. Start there.