Real estate comps (comparable sales) are recently closed MLS sales of homes a buyer would cross-shop against a subject property, weighed alongside active and pending competition. Under Fannie Mae's June 2025 Selling Guide, valid comps align on micro-market boundaries, recency, size, and UAD condition tier rather than raw price per square foot. Because off-market Zestimates carried a 7.21% nationwide median error rate in November 2026, MLS-connected AI comp tools score similarity and read listing photos so agents price with transparent evidence.

Pricing conversations rarely break down over math; they break down because the seller and the agent are looking at different comparable sales. Real estate comps (short for comparable sales) are recently closed MLS transactions of properties a real buyer would cross-shop against the subject home, evaluated alongside active and pending competition. When you select comps against a repeatable standard and show why each sale counts, the market carries the hard conversation instead of your opinion.
What makes a real estate comp valid?
A valid real estate comp is a closed sale that appeals to the same buyer pool as the subject property and requires minimal, market-supported adjustments. Under Fannie Mae Selling Guide B4-1.3-08 (updated June 4, 2025), a credible sales comparison starts with at least three closed sales drawn from the subject's market area, defined as the geographic region where most demand and competition sit.
Consumer blogs often reduce comping to a flat rule ("pull three to six sales within one mile"). Working listing agents and appraisers screen candidates across five stricter standards. Doing that screening by hand means exporting MLS sheets and clicking through dozens of photo galleries, whereas newer MLS-connected tools like Stairtop build and score the comp set from live MLS data in about a minute.
Micro-market boundaries, not radius circles. Start inside the same subdivision, tract, or school attendance zone before widening by distance. Crossing an arterial road, railroad track, or school boundary shifts buyer demand even a quarter-mile away.
Closed recency anchored to the contract date. Favor sales closed within the past 90 to 180 days, widening to 12 months only in thin markets. Under Fannie Mae Selling Guide B4-1.3-09 (updated June 4, 2025), market-condition time adjustments measure price movement from the comp's contract date to the effective date, not the closing date, because buyers negotiated 30 to 45 days before recording.
Matching property type, size, and layout. Compare detached homes only to detached homes, staying tight on finished living area, bed and bath count, lot utility, and build era. Fannie Mae B4-1.3-09 rejects flat rule-of-thumb square-foot adjustments when local sales show a different market reaction.
Aligned condition tier and verified concessions. Group comps using objective condition tiers instead of vague labels. Under Fannie Mae Selling Guide B4-1.3-06 (updated June 4, 2025), Uniform Appraisal Dataset (UAD) ratings run from C1 (new, never occupied) and C2 (fully remodeled to current standards) through C3 (well-maintained with selective updates), C4 (adequately maintained with dated finishes), C5 (significant deferred maintenance), and C6 (structural or safety defects). Check MLS fields for seller-paid rate buydowns or closing credits, which Fannie Mae B4-1.3-09 requires adjusting downward when they inflated the recorded sale price.
Active and pending competition. Consumer guides often advise sellers to ignore active and pending listings. Under Fannie Mae Selling Guide B4-1.3-07 (updated June 4, 2025), analyzing closed sales, contract sales, and active listings together is essential, especially in shifting markets. Closed sales show what buyers paid last month; active listings show what buyers can buy instead this weekend.
How do manual MLS comps, consumer AVMs, and AI comps compare?
Manual MLS filtering gives agents full control at the cost of 30 to 60 minutes of gallery clicking per appointment, consumer AVMs trade interior accuracy for instant estimates, and MLS-connected AI comp tools combine live transaction feeds with photo-based condition scoring.
In the traditional workflow, agents filter MLS sales by subdivision, square footage, and close date, export a spreadsheet or static CMA PDF, and click through photo galleries by hand to compare finishes. Manual MLS comping is thorough when you have time, yet it is hard to standardize across a team and easy for a defensive seller to dismiss as personal opinion. In Stairtop, live MLS feeds handle that filtering and photo review automatically, scoring each comp for similarity and summarizing its condition from listing photos so the agent walks in with transparent market evidence instead of a static printout.
Sellers usually arrive anchored to consumer AVMs. As of November 8, 2026, Zillow's published Zestimate accuracy data shows a 1.80% nationwide median error rate for on-market homes (measured right before pending, after the agent's list price and MLS details go live) versus a 7.21% median error rate for off-market homes. Redfin's published Estimate accuracy data (updated November 19, 2026) shows the same gap: 1.95% on-market median error versus 7.70% off-market, with only 35.79% of off-market estimates landing within 5% of eventual sale price. Because listing-appointment sellers are viewing an off-market estimate built from tax records and county-wide extrapolations, consumer AVMs miss unreported remodels, lot defects, and active competition.
Most software ranking for AI real estate comps today targets wholesalers and house flippers calculating After Repair Value (ARV) from county deeds rather than residential listing agents advising homeowners. Residential MLS-connected AI comp tools work differently: they ingest live MLS records, score candidate sales on location, recency, size, and feature similarity, and run computer vision over listing photos to flag interior condition differences. Automated models still require human oversight. The six-agency federal rule on AVM quality control standards (12 CFR Part 1222, effective October 1, 2025), announced by the Board of Governors of the Federal Reserve System, underscores why automated valuations need confidence controls, data-integrity safeguards, and testing. In thin rural markets or custom homes, an agent walking the property remains irreplaceable.
Dimension | Manual MLS Filtering | Consumer AVM (Zestimate / Redfin) | Investor ARV Tool | MLS-Connected AI Comps |
|---|---|---|---|---|
Primary data source | Local MLS sold, pending, active records | Tax records, deeds, county trends | Public deeds, distressed data, partial MLS | Live MLS feeds (closed, pending, active) |
Photo & condition awareness | Agent clicks galleries manually | Misses unpermitted interior updates | Repair-tier flip estimates | Computer vision reads listing photos and summarizes condition |
Similarity scoring | Agent filters manually | Black-box number or range | ARV weighting for investor offers | Transparent similarity score per comp |
Kitchen-table clarity | Static PDF; feels like opinion | Zero comp narrative | Built for cash offers, not retail sellers | Visual comps paired with market timing |
Key limitation | Slow; inconsistent across teams | High off-market error; misses nuance | Wrong lens for retail listings | Still needs agent walkthrough in thin or unique markets |
How do you score comps and spot a bad outlier?
Scoring comps by similarity and pairing each closed price with a photo-based condition note shows why a nearby high sale often does not belong in the pricing bracket. Consider an illustrative subject listing: a 2,050-square-foot, 3-bedroom, 2.5-bathroom 1998 two-story home in well-maintained, dated C4 condition (original oak cabinets, clean paint, intact roof), testing a $545,000 target list price.
Illustrative Comp (Distance) | Status & Timing | Price | Spread vs. $545K | DOM | Similarity Score | Photo-Based Condition Note | Verdict |
|---|---|---|---|---|---|---|---|
412 Maple Ridge Ln (0.2 mi) | Closed 28 days ago | $538,000 | -$7,000 | 19 | 94 / 100 | Original 1990s oak kitchen, updated LVP floors, clean yard (C4) | Primary closed comp |
438 Oak Hollow Dr (0.3 mi) | Closed 52 days ago | $549,000 | +$4,000 | 14 | 91 / 100 | Painted cabinets, quartz counters, original baths, minor trim wear (C3/C4) | Primary closed comp |
509 Cedar Creek Ct (0.4 mi) | Active listing | $555,000 | +$10,000 | 26 | 89 / 100 | Dated kitchen and baths matching subject; sitting with no contract (C4) | Active price ceiling |
385 Birchwood Way (0.2 mi) | Closed 41 days ago | $615,000 | +$70,000 | 9 | 68 / 100 | Full gut remodel: custom cabinetry, new roof, tiled baths, covered patio (C2) | Disqualified outlier |
Reading this illustrative four-comp set against Fannie Mae's condition definitions narrows the range immediately. The two highest-similarity closed sales bracket value between $538,000 and $549,000, while the active C4 listing at $555,000 proves buyers are stalling above $550,000 for original finishes. Although 385 Birchwood Way sits two blocks away and closed in nine days, its C2 gut remodel drops its similarity score to 68 out of 100 and puts it in a different buyer tier.
Building that side-by-side comparison by hand means opening dozens of MLS photo tabs and typing notes into a spreadsheet or Cloud CMA PDF, which leaves room for sellers to question your judgment. In Stairtop, each curated comp displays its price spread against your target price, days on market, similarity score, and a short photo-generated condition note side by side, so a remodeled outlier disqualifies itself on screen.
How do you handle a seller's cherry-picked Zillow comp?
To handle a seller's cherry-picked Zillow comp, acknowledge the higher sale immediately, put its listing photos and condition notes beside closer comps on screen, and anchor the seller's decision to active competition. Sellers trust evidence they can verify visually.
"I pulled 385 Birchwood Way too, because it's two blocks over and $615,000 jumps off the page. In the listing photos, Birchwood had a down-to-the-studs kitchen, new tile baths, a new roof, and a covered patio, so it traded in a remodeled tier. Right next to it, 509 Cedar Creek has the same floor plan and original kitchen as your home, listed at $555,000, and buyers have walked past it for 26 days with no offer. Based on the closer closed sales at $538,000 and $549,000, launching at $545,000 puts us in front of active buyers this weekend instead of helping Cedar Creek sell first. Would you rather price in the $538,000 to $549,000 band where homes went pending in two to three weeks, or test higher knowing buyers are stalling at $555,000?"
If the seller pushes back ("We don't need $615,000, what if we test $575,000 first?"), stay calm and tie asking price directly to time on market:
"We can list at any number you choose, and my job is to show you what happens at $575,000 before we commit. At $575,000, buyers searching online see Cedar Creek at $555,000 for the same unrenovated kitchen and compare your home against updated properties near $580,000. If you want to test above $550,000, let's put a 14-day review date in writing so we adjust before the listing goes stale."
A static CMA printout cannot answer "what happens if we test $575,000" without recalculating by hand. Inside a Stairtop Seller Presentation, dragging the list price on the live time-to-offer curve updates predicted days to pending (within about 6 days of actual market outcomes on average) and estimated seller net proceeds in real time, turning a debate over opinion into a visible trade-off.
How do you put scored comps to work at a listing appointment?
You put scored comps to work by opening the live comp set and pricing curve on one screen so the seller sees the market gap for themselves. Picture a Tuesday afternoon appointment where the seller already has a neighbor's remodeled sale printed from Zillow. Before leaving the office, you enter the subject address into a Stairtop Seller Presentation and pull the scored comps in about a minute. At the kitchen table, each comparable sale shows its similarity score, days on market, and photo-based condition note beside the live time-to-offer curve and estimated seller net proceeds. When the seller asks to test $30,000 higher, you drag the list price up together and watch the predicted pending date stretch out while the comp gap stays visible. Across Stairtop Seller Presentations, 93% of listings sold faster than the market median.
Solo agents can preview a Seller Presentation in about a minute at app.stairtop.com/try, and team leaders or managing brokers standardizing comp presentations across an office can request a brokerage demo.
Frequently asked questions
How do I look up real estate comps?
Licensed agents pull real estate comps directly from their local Multiple Listing Service (MLS), which captures closed sale prices, contract dates, interior listing photos, days on market, and seller concessions. Homeowners without MLS access can filter by recently sold properties on public portals like Zillow or Redfin and verify deeds in county recorder offices, though public records lag closings and omit interior condition and concession details.
What do real estate agents use for comps?
Real estate agents rely primarily on live MLS sold, pending, and active listing data to build a Comparative Market Analysis (CMA) or Seller Presentation. Many agents pair MLS exports with CMA software, public tax records, or MLS-connected AI comp tools that score property similarity and summarize condition from listing photos before the listing appointment.
Is AI actually good at real estate comps yet?
General AI chatbots are unreliable for real estate comps because they lack direct MLS feeds and can fabricate sale prices, while investor ARV tools are tuned for distressed flips rather than retail listings. Specialized MLS-connected AI comp engines are strong at curating and scoring residential comps in active suburban and urban markets, especially when they evaluate listing photos, though agents still need to verify interior walkthrough details and thin rural markets.
Why are Zillow comps or Zestimates different from MLS comps?
Consumer portals rely heavily on tax assessor records, prior sales, and broad county-level trends for homes that are not yet listed. As of November 2026, Zillow reported a 7.21% nationwide median error rate for off-market Zestimates and Redfin reported 7.70% for off-market Redfin Estimates, because automated consumer models cannot see unreported remodels, street-level defects, or seller concessions hidden inside MLS transaction history.
How far back should real estate comps go?
In active residential markets, agents prioritize comparable sales that went under contract and closed within the past 90 to 180 days so prices reflect current mortgage rates and buyer demand. Under Fannie Mae Selling Guide B4-1.3-08, closed comps within the last 12 months are standard for appraisals, and older sales remain acceptable in thin or rural markets when supported with market-condition time adjustments.