Take one CLO portfolio and ask two questions. What does the indenture say its Fitch recovery rate is? And what does Fitch say? On the day a deal was signed, those answers matched. For any deal still carrying earlier Fitch language, they no longer do. That gap is the issue.
On June 1, 2026 Fitch replaced its CLOs and Corporate CDOs Rating Criteria, superseding the version dated July 21, 2023. The changes focus mostly on recovery, and coverage since has focused on upgrades. That is the right headline for ratings. For documents, it is the wrong one.
An indenture tests recovery using its own definitions, frozen on the day it was signed. Fitch models the deal using current criteria. Where the two have drifted apart, the manager’s compliance tests and Fitch’s model measure different things. This comes to a head when the deal goes back to Fitch: at a reset, refinancing, or upgrade review. US CLOs saw about $164bn of resets (down about 15% year on year) and $99bn of refinancings (up about 30%) in the first eight months of 2026, so deals are returning to Fitch quickly.
What changed, line by line
Five changes matter for documents. Each one below is taken from the two criteria reports side by side, not from secondary coverage.
| Area | July 2023 criteria | June 2026 criteria |
|---|---|---|
| Asset recovery | Fitch Recovery Ratings (RR1 to RR6) or recovery estimates used where available; standard categories only as a fallback | Recovery Ratings no longer used in CLO analysis; every asset gets a standard assumption by seniority and jurisdiction |
| US middle-market senior secured loans | Own “Strong recovery MML” category: 65% recovery factor in the Fitch WARR | Category deleted; treated as “Strong recovery”: 75% |
| Deals with no Fitch WARF test | ‘CCC’ bucket maximized in the stressed portfolio | Same, plus a WARF stress of at least 1.5 points over the indicative portfolio |
| ‘CCC’ overcollateralization stress | Multiples of 2.5x (‘BBsf’) and 3.5x (‘Bsf’) only | 1.5x (‘Asf’), 2.5x (‘BBBsf’), 3.5x (‘BBsf’), 4.5x (‘Bsf’) |
| Recovery lag | US 12 months; Europe 18 months at ‘Asf’ and above, 12 below | US 6 months; Europe 12 months |
The middle-market change runs through the whole model, not just the WARR scale. At every rating, the old middle-market assumptions sat 5 to 10 points below standard senior secured loans; they now match.
The ‘CCC’ change is the one most readers will miss. In 2023 the haircut on excess ‘CCC’ assets was modeled only at ‘BBsf’ and ‘Bsf’. It now extends up to ‘Asf’, and the ‘Bsf’ multiple of 4.5x exceeds the 4.0x historical average Fitch cites.
Two quieter edits also affect documents. The WARR-to-recovery interpolation grid used for Fitch Test Matrices was replaced: a 62.5% WARR covenant now maps to a 57.5% ‘BBBsf’ rating recovery rate, up from 55.25%. The appendices were renumbered, so an indenture citing “Appendix 10” for the grid now points to a different section.
The net effect on ratings is positive. Fitch’s own estimate, as reported by Janus Henderson, is that 5% to 15% of its CLO ratings could be upgraded, mostly by one or two notches.
Same loan, two recovery numbers
The mismatch runs in both directions, and which way it runs depends on what the deal owns. That is the uncomfortable part: you cannot tell from the vintage alone.
Many indentures copy Fitch’s recovery-factor table into the Fitch Recovery Rate definition. A deal drafted to the 2023 criteria scores each US senior secured loan on its Recovery Rating where one exists. Fitch now scores the same loan at a flat 75%, unless its analyst moves it to a lower category.
| US senior secured loan | 2023 factor (in many indentures) | June 2026 factor (Fitch today) | Covenant vs Fitch |
|---|---|---|---|
| Rated RR1 | 95% | 75% | Looser by 20 points |
| Rated RR2 | 80% | 75% | Looser by 5 points |
| No Recovery Rating, public issuer | 75% | 75% | Aligned |
| Rated RR3 | 60% | 75% | Tighter by 15 points |
| Middle-market, no public rating | 65% | 75% | Tighter by 10 points |
| Rated RR4 | 40% | 75% | Tighter by 35 points |
Factors are the Fitch WARR recovery factors (‘BBsf’ stress) in Appendix 6 of each criteria report.
A portfolio heavy in RR3 or middle-market names has a covenant WARR below what Fitch calculates. The manager faces a tighter test than the agency applies and is giving away trading room for nothing.
A portfolio heavy in RR1 and RR2 names has the opposite problem. The covenant says the deal is passing comfortably. Fitch’s model at the next reset may say the cushion is thinner than reported.
Neither case is a rating event by itself. Both are surprises, and both arrive at the moment the manager has least room to absorb one.
When it bites
The recovery changes apply at every Fitch review, including annual surveillance on the actual portfolio. The sharpest exposure comes when Fitch rebuilds its stressed portfolio, built to the deal’s covenant limits. That is what it does for new ratings and upgrades of CLOs still reinvesting.
That gives three triggers:
- Resets. A reset is a new rating, run on the June 2026 criteria from scratch.
- Refinancings with covenant changes. Refinancings that bring material credit-negative covenant changes face the same rating tests as new ratings. A plain spread refinancing may not trigger a full re-run, so the trigger depends on the covenant changes, not the refinancing alone.
- Upgrade reviews. For a reinvesting deal, an upgrade is tested against the stressed portfolio, not the actual one, so the trigger is the upgrade review itself.
The new WARF rule sharpens this for one group. Fitch notes that its own tests are often absent from US BSL CLOs where another agency rates the subordinated notes. For those deals the criteria now say Fitch will stress the WARF value by at least 1.5 points from the indicative portfolio.
Fitch’s rule of thumb is that each WARF point adds roughly 1% to the rating default rate. On a deal priced to a thin cushion, 1.5 points is not decoration.
The timing is not theoretical. JPMorgan expects $422bn of CLOs to exit non-call periods in 2026, and US resets and refinancings totaled about $263bn by August 31. Every Fitch-rated deal among them goes back through Fitch’s model when those triggers hit, while its indenture still holds the definitions it was signed with.
What the documents show
We reviewed Fitch-rated US CLO documents in the Semeris library: 38 middle-market deals, read clause by clause as of October 14, 2026, and a random sample of 50 BSL deals that Fitch rates and that carry Fitch portfolio tests. The BSL sample was drawn with a fixed seed from the library as of August 31, 2026, stratified by the date of the latest document restating the Fitch definitions. Where a deal’s latest document is a marked-up amendment, we read the amended text, not the deleted text. The library may include deals that have since been called or redeemed.
Middle-market deals are the clearest case. Of 38 US middle-market CLOs listing Fitch as a rating agency, at least 32 contain a Fitch recovery-rate definition, and 29 of those could be read and classified in full. All 29 still score assets on Fitch Recovery Ratings (RR1 to RR6), which Fitch no longer uses in CLO analysis.
BSL deals not restated since June are in the same position. 40 of the 50 BSL deals still score assets on Recovery Ratings, and all 40 were last restated before June 1, 2026. All 10 restated since have moved to the asset-type table the current criteria use, which scores senior secured loans at a flat 75%. A deal’s definition stays as signed until the deal is restated or Fitch notifies a different rate.
Few of those definitions can be updated by the deal itself. Of the 29 middle-market definitions, 25 are fixed: they change only if Fitch notifies a different rate, or if the indenture is amended. The other 4 let the issuer swap in a replacement definition once Fitch confirms in writing that no rating will be immediately cut or withdrawn, or waives its review. In 44 of the 50 BSL deals, the recovery rate changes if Fitch notifies a different one. Otherwise, the definition stays as written until the indenture is amended.
The Fitch Test Matrix is easier to switch than to change. Of 30 middle-market deals with a matrix, 12 say how its values can be changed. The routes run from a Fitch-supplied replacement matrix that needs no noteholder consent, to a Controlling Class majority, to separate majorities of the Class A and Class B notes. The other 18 don’t address it. BSL deals typically let the manager move between existing matrix cases on notice. Changing the values themselves varies widely:
- 9 of 50 let the manager modify the matrix subject to rating-agency confirmation.
- 13 name the matrix in their amendment provisions. Most need a noteholder vote, usually a Controlling Class majority, and two also need a majority of the Subordinated Notes. Three need no vote at all: one only on notice to Fitch, at a refinancing, re-pricing or additional issuance.
- 28 don’t address it, so a change falls back on the general amendment provisions.
The same change can take a letter to Fitch in one deal and a two-class noteholder vote in another.
Some deals have no Fitch WARF test at all. This is common in older BSL deals, where another agency’s tests govern the portfolio. These are the deals the new 1.5-point stress is written for, and their managers may not see it until the reset model comes back.
None of this is visible from a trustee report. It sits in definitions and amendment provisions, a few hundred pages into each deal.
Before your next reset
The lesson is simple: the criteria changed once on June 1, but each indenture changes only when amended. Five questions settle where a deal stands and what will trigger a review:
- Does the Fitch Recovery Rate definition still score assets on Recovery Ratings or the 2023 recovery tables?
- Does the portfolio lean toward RR1 and RR2 names (covenant looser than Fitch) or RR3, RR4 and middle-market names (covenant tighter)?
- Is there a Fitch WARF test? If not, budget for the 1.5-point stress at the next new rating or upgrade.
- How does the definition change: by substitution once Fitch confirms no downgrade, only on notice from Fitch, or only by amendment?
- What consent does a change to the Fitch Test Matrix need, and does the reset timetable leave room to get it?
The immediate win is a reset with no modeling surprises. The longer-term win is a document set that tracks the agency, not a snapshot of it as it stood in 2023. That is the point to keep in view before the next reset.
Our library holds more than 1,100 Fitch-rated US BSL deals, and each of these five questions can be answered from the documents at clause level. If you would like to know which side of the line your deals sit on before Fitch tells you, talk to us.
Sources
- Fitch Ratings, CLOs and Corporate CDOs Rating Criteria, June 1, 2026: Rating Determination, Portfolio Credit Quality, Timing of Recoveries, ‘CCC’ overcollateralization stress, Appendices 4 and 6.
- Fitch Ratings, CLOs and Corporate CDOs Rating Criteria, July 21, 2023, as filed in Fitch’s Form NRSRO exhibit: Appendices 4, 6 and 10.
- Janus Henderson, Demystifying recent rating agency methodology changes for CLOs, June 16, 2026.
- Valuation Research Corporation, Structured Products Market Trends: Q3 2026, data to August 31, 2026.
- PitchBook LCD, 2026 US CLO Outlook, citing JPMorgan.
- Semeris document library: 38 US middle-market CLOs and a random sample of 50 Fitch-rated US BSL CLOs.