What is the Information Quality Standard (IQS)?

The IQS is a publicly accessible standard that grades every article on its sourcing, framing, and conflicts of interest — so you can see what you're actually reading, instead of trusting a whole outlet's reputation. The scoring methodology is proprietary and confidential; the standard and its results are open to everyone.

How does Lens score an article?

Lens analyzes each article across multiple signals — language and framing, source profile, and conflicts of interest — and produces a transparent 0–100 score with a sentence-by-sentence breakdown of what affected it.

How is Lens different from a fact-checker?

Lens scores at the article level using one consistent publicly accessible standard rather than rating whole publications, and shows exactly which sentences and factors drove the score.

Lens offers a free tier for everyday readers, with paid tiers for deeper analysis and professional or institutional research.

Work with us›Join the programIn development · design partners now

Design Partner Program · AI Platforms and Agents

Eliminate hallucination through auditability.

Every AI product cites the passage it built the answer from. That catches invention. It does not catch a real citation from a bad source. Lens closes that gap at the document and claim level, on in-house models, with the evidence attached to every flag.

Apply for a slot

example.com/articlescored

Excerpt · flagged passagesThe department’s new screening rule reads as a naked cash grab dressed up in public-health language. Meanwhile the parent group filed disclosures showing it quietly increased its holdings in a firm that competes head-on with the diagnostic under review, a fact the piece never attributes to any named source.

64/100

composite

Sample output · every flag carries its reasonevidence attached

01Why this program exists

The citation catches invention. It does not catch a bad source.

Retrieval traces the answer back to a real passage. It does not audit what that passage is worth. Lens does, per document and per claim, with the reason exposed for every score.

49.9%

of new English-language online articles are AI-generated. Your retrieval index is drinking from that river.

Graphite, Q1 2026

13.6%

mean rate at which AI-judge verdicts flip between two identical runs. Grounding a bad source in a citation just launders it.

arXiv, 2026

third-party document-level standards your framework can call today. That is the seat this program opens.

The gap Lens fills

02What partners get

Access, price, and a seat at the roadmap.

Early access to the source-scoring API.

Full evidence-attached scoring output, delivered through the integration surface that fits your stack.

Direct roadmap input.

Standing working session with the engineering team. What ships next is shaped by what your framework actually calls, not by a survey.

Named engineering contact.

A named engineer on the Lens applied research team, same time zone as your integration lead. Not a shared inbox.

One integration path, chosen with you.

Multiple integration surfaces sized to your stack. Wired end to end with your team on the shape that fits.

Partner pricing through the program term.

A per-call rate held for the duration of the cohort. Volume ceiling agreed in advance. No surprise metering.

Public co-announcement, optional.

Some partners prefer quiet through the term. The program does not require the logo.

03What we ask back

A working integration and honest feedback.

Real traffic through the API.

A running integration seeing production shape and volume, not a proof-of-concept parked on a branch.

A standing working call.

Thirty minutes on the agreed cadence. What was called, what worked, what did not, what to change. Engineering team on every one.

Permission to learn from calls in aggregate.

Anonymized. Distributional. Never a specific user query. Terms written into the program agreement.

An early integration reference.

A call with the next cohort or a quote after program close. Both optional.

04Program shape

Small cohort. Terms negotiated per partner.

Cohort

Term

Admission

SizeA small number of slots. Each partner served in depth, not scale.

LengthSet per partner in the contract. Length is negotiated against the integration depth agreed on the technical call, not against a template.

CadenceRolling. New partners onboard as slots open, not on a fixed calendar.

FitAgent frameworks, retrieval systems, browser or search products, editorial tools built on LLMs.

MilestonesSet at contract. Integration, first data review, and roadmap sign-off scheduled against your program plan.

CriteriaReal users on the other side. A retrieval or authoring surface. A named engineering owner.

05What we will not do

Four lines named up front.

Political filtering.Lens is a quality and integrity measurement. There is no lean signal in this product and there will not be. If your product needs one, this is not the layer.

Domain allowlists.Lens scores the document. A good domain does not raise a bad article, and a weak domain does not sink a strong one. If your product ships an allowlist today, verification runs alongside it, not underneath it.

Third-party AI in the serving path.Scoring runs on in-house models on Lens infrastructure. No other model sits between your call and the score.

Verdicts on truth.Lens returns a measurement with the evidence attached. You set the threshold. Your product owns the call.

06Onboarding

Application to production, on your integration’s own timeline.

Apply

Draft an email with the two sentences below and the one integration surface you have in mind.

Scoping call

Thirty minutes with the applied research team. What you are building, where the check would sit, what would count as working.

Access provisioned

Keys issued and your integration surface stood up. First calls verified end to end with your team.

Live in your product

First real calls. First working session. First look at the roadmap you now have a seat at.

07Questions we get

Direct answers.

How does pricing work?

Terms are negotiated per partner and written into the program agreement. Details on the intro call.

Can we run this on our own infrastructure?

Private deployment is handled on the enterprise track. Start on the enterprise page.

What is the term?

Set per partner in the program agreement, sized against integration depth. Two-weeks notice on either side is the default exit clause.

Who reads the application?

The engineering team. Every application, read within one business day.

08Apply

Two sentences and a surface.

Program application

Fill this in and it goes straight to the Lens team. Every application is read within one business day.

Not the right fit right now?

Two other lanes are open. Send a campaign log and get a pilot audit back, or talk about verification inside a regulated stack.

Request a pilot audit Talk to us