What LP Due Diligence on GPs Actually Looks Like in 2026

What track record review actually uncovers

Most GPs enter LP meetings believing their track record is their strongest asset. It often is. But the way LPs evaluate that track record is not the same as the way the GP presents it, and the gaps between those two interpretations are where allocations are lost.

The first question is attribution. For multi-partner firms, LPs want to know which partners sourced, underwrote, and managed each investment. They want a deal-by-deal account: sourcing role, investment thesis at entry, how the thesis evolved during the holding period, and the realized outcome. They are not interested in anecdotes about how the team "worked together" on every deal. That framing is precisely what makes attribution analysis difficult. When a firm's best-performing vintage was led by a partner who has since left, the track record does not transfer.

For buyout funds, academic research on performance persistence offers a sobering baseline. Post-2000 data shows "little or no evidence" of persistence at the time of fundraising, meaning that a GP's prior fund quartile ranking has limited predictive value for the next fund. Only bottom-quartile persistence remained statistically significant, which is a polite way of saying that bad managers tend to remain bad. LPs who understand this use track records as one input among several rather than as a primary decision criterion.

The second question is the interim fund warning. Research on private capital markets has found that the interim results of a predecessor fund (the valuations showing in the portfolio when the next fund is being raised) have very little predictive value for final outcomes. Unrealized marks can be written up or down with considerable discretion under current accounting standards, and GPs who raise successor funds while their prior fund is still in the holding phase have an obvious incentive to maintain optimistic interim valuations. LPs who have experienced this dynamic discount interim marks significantly and focus on realized multiple and DPI from fully or nearly fully liquidated positions.

The third question is scale. A GP that generated 25% net IRR deploying 200 million USD is not necessarily capable of generating the same return deploying USD 800 million. Deal size increases, competitive dynamics change, and the sector coverage required to find enough qualified investments at larger fund sizes is different in kind, not merely in degree. LPs ask explicitly: how will your strategy change at this fund size? What is the largest single investment you anticipate making, and how does that compare to the largest investment in the predecessor fund?

LP diligence workstreams increasingly use AI to process and compare track record data, LPA provisions, and DDQ responses across multiple manager submissions simultaneously.

Reference calls: who LPs actually contact

The reference call process is the part of LP due diligence most GPs manage least effectively. This is partly because GPs do not control it, and partly because they underestimate the range of people LPs contact.

LPs draw references from six sources: current and previous LPs who have committed to prior funds; senior leadership of portfolio companies, both current and exited; investors who were presented with the fund opportunity and chose not to commit; service providers including administrators, auditors, and outside counsel; background checks on unknown service providers; and past employees of the GP firm. That last category is the one most GPs overlook. A former associate who left the firm two years ago and is now at a different shop is likely to give a candid account of investment process quality, team dynamics, and how the GP actually behaved with portfolio companies under stress.

LPs also use their networks to identify references the GP did not provide. They ask their own contacts in the industry who else has worked with or evaluated this manager. Those second-degree references are often more candid than the references GPs submit directly, for an obvious reason: they have not been pre-selected for their willingness to say positive things.

The questions asked on reference calls follow a consistent pattern. Does the GP's reporting quality and frequency match the commitments made during fundraising? Has there been any situation where GP and LP interests diverged, and how did the GP handle it? Has the GP requested any amendments to the LPA since closing, and what were the circumstances? Would the LP commit to a successor fund from this manager, and why or why not?

That fourth question is the most informative. An LP who says they would re-up but cannot articulate a clear reason is giving a different signal than one who says they would re-up and explains why their experience matched expectations in three specific ways. An LP who declines to answer, or who becomes evasive, is giving the clearest signal of all.

The AI due diligence agent structures document review outputs into consistent formats, allowing LP teams to compare GP responses and financial data across multiple fund submissions in the same analytical framework.

Key person risk as an underwriting question, not a contract clause

Every fund LP agreement includes key person provisions: if a defined set of senior investment professionals devote less than a specified percentage of their time to the fund, LP rights are triggered. This clause is standard. What is less standard is the rigor with which LPs think through key person risk before making a commitment, as opposed to relying on the clause to protect them after a problem arises.

In practice, LPs treat key person evaluation as a qualitative underwriting question. They want to know whether the firm's investment process is person-dependent or system-dependent. A firm that sources deals through one partner's relationships, underwrites through another's judgment, and manages portfolio companies through a third's operational instincts has concentrated risk in three individuals. A firm with documented investment processes, shared deal coverage, and an associate class that has been promoted to principal roles has distributed that risk across a larger organization.

The alignment question reinforces this. The average GP commitment to a fund is 4.8% of fund size, according to a 2021 survey of 219 general partners. First-time managers often exceed this, contributing more than the requirement as a signal of conviction. But the distribution of that commitment within the partnership matters as much as the total. If one founding partner holds 90% of the GP commitment and the rest of the team holds 10% combined, the retention incentive structure for non-founding partners is weak, and the key person risk is concentrated regardless of what the LPA says about it.

LPs who discover that key person definitions in the LPA are narrow, covering only the founding partner rather than the broader investment team, sometimes negotiate expanded definitions during LPA review. This negotiation, combined with management fee offsets and LPAC representation, represents the governance lever LPs can pull before the relationship begins. The no-fault divorce provisions that 18% of funds saw exercised in 2025 represent what happens when those levers were not sufficient.

Fund economics: when management fees signal misalignment

The standard private equity fee structure (2% management fee, 20% carried interest above an 8% preferred return) is a starting point, not a norm. LPs evaluate fund economics for their absolute terms and for what they reveal about the GP's incentive structure and financial dependence on the fund.

The concern with management fees is not that they exist. They exist to cover the operating costs of running the fund: salaries, office expenses, travel, legal, and administrative overhead. The concern is when management fees substantially exceed those costs, which turns the fee structure into a profit center in its own right. A GP that is comfortable generating significant income from fee streams regardless of investment performance has a weaker incentive to prioritize carried interest, which is the alignment mechanism the whole structure is supposed to depend on.

LPs who are examining a larger fund from a manager that previously ran a smaller fund look at this question carefully. As fund size grows, management fee revenue grows proportionally, while the cost of running the fund grows less than proportionally. The economics improve significantly for the GP. Whether those improved economics come at the expense of LP alignment depends on whether the carry structure has been updated to reflect the changed economics, and whether the team that will earn that carry is the same team that built the prior track record.

Some LPs are open to non-standard economics when they believe superior results justify them: a higher carry share in exchange for a lower management fee, or back-ended management fees that increase only after a certain deployment threshold. What raises flags is when economics are non-standard in the GP's favor without a corresponding explanation of what the LP receives in exchange.

How AI document analysis is changing LP due diligence

The ILPA DDQ 2.0 alone runs to 75 pages in its complete form. Add a fund prospectus, a track record document, five years of audited financial statements, and a draft LPA, and an LP is managing several hundred pages of structured and semi-structured text before the first management presentation. The information is present. Extracting it consistently, and making it comparable across multiple GP submissions being reviewed simultaneously, is where AI document tools have begun to change the workflow.

The LPA analysis use case is illustrative. LP legal teams traditionally review limited partnership agreements manually, flagging provisions that deviate from market standard: key person definitions, LPAC composition, no-fault divorce thresholds, management fee offsets, and waterfall structures. An AI LPA analysis agent can run that review against a defined extraction schema, surfacing the specific clause language for each defined provision, flagging deviations from the LP's preferred terms, and producing a structured comparison across multiple LPA submissions. The human lawyer still makes the judgment about whether to negotiate; the AI compresses the time required to understand what is in front of them.

Track record document review follows the same pattern. A GP submits a track record in a format that reflects their own preferences: which investment to present first, how to define the relevant universe, how to handle the fund that underperformed. The LP receives it in whatever form the GP chose. An AI extraction agent applies a consistent schema across all GP track record submissions: realized and unrealized IRR, MOIC, DPI, holding period, deal size, and whether realized investments account for more or less than half the total capital deployed. The schema does not eliminate judgment. It ensures the comparison is made on equivalent terms.

AI document tools applied to LP due diligence reduce the time-to-insight gap between receiving a GP submission and completing structured analysis across investment, organizational, and operational workstreams.

For GPs, the operational implication runs in the other direction. If LPs are applying extraction schemas to DDQ submissions and track record documents, the quality of structured data in those documents becomes more consequential. A DDQ completed with consistent formatting, precise figure conventions, and unambiguous answer structures will fare better in an automated extraction workflow than one built around narrative paragraphs that require manual interpretation. This is an argument for GPs to think about their submission documents not only as investor communication but as machine-readable data.

The AI due diligence agent supports both sides of this workflow: LPs using it to process GP submissions at scale, and GPs using it to review their own documents before submission, identifying where structured data is missing or where claims require clearer evidence. The role of AI in due diligence is not to replace the judgment calls that determine whether to allocate. It is to ensure those judgment calls are made on the basis of complete, consistently extracted information rather than on whatever the GP chose to highlight and whatever the LP's analyst happened to notice.

What LPs are asking that they were not asking five years ago

The questions LPs ask today reflect regulatory changes, structural shifts in private markets, and a decade of lived experience with what actually goes wrong in GP-LP relationships.

On the regulatory side: the SEC's 2026 examination priorities explicitly include valuation methodology for illiquid assets, management fee calculation and offset practices, and disclosure of conflicts around fund-level credit lines, adviser-led secondaries, and affiliated service providers. LPs now ask GPs directly: have you been subject to an SEC examination in the last five years? What was the scope? Were there any deficiency letters? These questions used to feel aggressive. They now feel standard.

On the structural side: subscription credit lines have become common, and LPs have become more sophisticated about their effect on reported IRR. The ILPA DDQ 2.0 updated Section 14.0 to require GPs to report track record returns both with and without the effect of subscription credit facilities. LPs who receive only the facility-inclusive figures now ask for the gross-up explicitly. A fund that reports 28% IRR with a credit line and 21% IRR without it is not necessarily misrepresenting its performance, but an LP who allocates based on the 28% figure without understanding the difference has made a more leveraged bet than they may have intended.

On operational infrastructure: ESG integration questions have migrated from optional addendum to core DDQ. Over 65% of new institutional commitments in 2024 required ESG reporting as a condition of investment. LPs ask whether a GP has an ESG policy and whether ESG considerations are integrated into investment screening, and whether portfolio company reporting includes quantitative environmental and social metrics. GPs without processes to support those commitments face follow-up questions they cannot answer with a policy document.

The cumulative effect of these additions is that LP due diligence now covers ground that investment-focused GPs did not historically think of as diligence territory. A buyout manager who has spent thirty years thinking about investment strategy and deal execution may be less prepared to discuss cybersecurity incident response protocols or attribution methodology than they are to discuss market entry multiples. That gap is exactly what thorough LP due diligence is designed to find.