PulsCar analysis of NHTSA public data, September 2026 Clicking noises: 4,898 owner complaints, 1,228 manufacturer communications Licensed CC BY 4.0. Queries run 18 September 2026. -------------------------------------------------------------------------- SOURCE FILES -------------------------------------------------------------------------- NHTSA ODI complaint flat files COMPLAINTS_RECEIVED_2015-2019, _2020-2024, _2025-2026 -- https://static.nhtsa.gov/odi/ffdd/cmpl/ (dictionary CMPL.txt) 777,040 distinct vehicle complaints (ODINO, product type V) received 1 January 2015 to 8 September 2026. NHTSA manufacturer communications (TSB) flat file FLAT_TSBS.zip. NHTSA recall flat file FLAT_RCL.zip. Manufacturer bulletins downloaded from static.nhtsa.gov, listed at the end. NOTE ON THE EARLIER NUMBER IN THIS ARTICLE. A previous pass (13 September 2026) counted 1,221 click complaints across 26 nameplates, model years 2015-2022, through NHTSA's public API. This pass replaces that with the whole complaint file -- every make, every model year, eleven years -- which is why the counts are several times larger. Both numbers are correct for their own scope; the API figure is kept in the article where it is labelled. -------------------------------------------------------------------------- WHICH COMPLAINTS COUNT (the clicking-noise set: 4,898) -------------------------------------------------------------------------- STEP 1 -- full-text search (FTS5) for the whole words click | clicks | clicking | clicked 7,828 complaints matched. STEP 2 -- three exclusions. (a) NOT A NOISE. A click mention counts only if a noise word -- noise | sound | hear | heard | hearing | audible | rattle | tick -- appears within 90 characters of it. This removes narratives where something "clicked into place," a switch clicked, or a control was clicked. (b) A CLICK THAT IS NOT THE CAR. Occurrences whose 180-character window names a gas pump, a fuel nozzle, a seat belt or buckle, a turn signal or blinker, a mouse, a website or an online form are not counted. Web-form phrasings ("click here", "click on the", "one-click") are dropped at the occurrence level before anything else. (c) NOT A PASSENGER VEHICLE. Heavy-truck, bus, RV, motorcycle and engine manufacturers were dropped (Freightliner, Kenworth, Peterbilt, Mack, International, Western Star, Hino, Blue Bird, Thomas Built, Winnebago, Thor, Jayco, Keystone, Newmar, Tiffin, Coachmen, Grand Design, Entegra, Dutchmen, Prime Time, Crossroads, Spartan, Gillig, New Flyer, Autocar, Starcraft, Alliance, Harley-Davidson, Ducati, Yamaha, Kawasaki, Suzuki Motorcycle, BMW Motorrad, Cummins, FCCC, Altec). 213 complaints were dropped as web-form-only, 9 as heavy or motorcycle makes, and 2,708 for having no noise framing or only a non-vehicle click. 4,898 remain. This is the denominator for every percentage below. SAMPLE READING (required before publishing any keyword count) Thirty randomly drawn matches from the 7,606 that survived the web-form and make filters were read in full BEFORE the noise-framing rule was applied. Twenty-seven described an audible clicking from the vehicle. Two were clicks that are not noises in the sense this article means -- a fuel pump nozzle clicking off during refuelling, and a door lock that clicked without locking. One was the ABSENCE of the turn signal's click. That is what motivated exclusions (a) and (b): FALSE-POSITIVE RATE BEFORE THE NOISE RULE, 3 IN 30, ROUGHLY 10%. -------------------------------------------------------------------------- GROUPS (click-complaints-by-group.csv) -------------------------------------------------------------------------- A complaint is counted in a group when its narrative contains one of that group's phrases. Complaints matching two groups are counted in both, so the groups do not sum to 4,898. Steering or turning : a click word and a turn/turning/steering word within 90 characters of each other, in either order, with no sentence boundary. All other groups: a straight phrase list, printed in the article and in the code comments of the query. Largest groups: steering or turning 2,177 (44.45%); dashboard, HVAC or actuator 623 (12.72%); brakes or wheel area 602 (12.29%); battery named 458 (9.35%); click with a no-start or no-crank 358 (7.31%); starter or solenoid 169 (3.45%); CV joint, axle or half-shaft 147 (3.00%); heat shield or insulator 7 (0.14%). -------------------------------------------------------------------------- RESULTS -------------------------------------------------------------------------- Outcome flags (share answering Y, of 4,898): crash 1.65%, fire 1.53%, towed 3.59%, 61 injuries, 0 deaths. All-complaint baseline (777,040): crash 4.37%, fire 1.82%, towed 5.11%. A clicking-noise complaint is markedly LESS likely than the average complaint to carry a crash or tow flag. Two subgroups break that pattern: the brakes-and-wheel-area group (crash 4.98%, three times the set) and the oil-pressure group (towed 15.09%, four times the set on 53 complaints). Mileage (MILES field, 0 < miles < 400,000; 2,712 complaints report one): median 73,000; quartiles 40,000 and 107,341. Baseline median 70,000. No fault found: 131 of 4,898 (2.67%) say a shop could not duplicate, reproduce, find or locate the noise. THE TURNING CLICK (turning-click-what-owners-named.csv). Of the 2,177 complaints tying a click to turning or steering: steering column, rack, gear, EPS or MDPS named .... 578 (26.6%) steering sticks, binds, stiffens, locks or jerks .. 245 (11.3%) CV joint, axle or half-shaft named ................ 194 (8.9%) brake or wheel area ............................... 159 (7.3%) suspension arm, joint, strut, bar or tie rod ....... 70 (3.2%) no part named at all ............................. 1,131 (52.0%) Steering-gear mentions outnumber CV-axle mentions roughly three to one. LIMITATION: these are the words owners wrote, not confirmed diagnoses, and the set is dominated by two clusters (below). It does not establish how often a turning click IS a CV joint. What it does establish is that a turning click is commonly reported as a steering-side problem, which is enough to make "clicking while turning equals CV joint" an unsafe default. TWO CLUSTERS DOMINATE THE TURNING GROUP Hyundai Sonata: 413 click complaints, 219 of them model year 2011; 37.0% name a steering column, coupler, rack, gear, EPS or MDPS; 1 complaint (0.2%) names a CV axle; median 80,000 miles. Hyundai TSB 17-ST-002 covers exactly this part on exactly these years. Honda Prologue: 335 click complaints, 287 of them model year 2024; 51.3% name a CV joint, axle or half-shaft; 0.9% name steering hardware; median 5,786 miles across the 14 complaints that state a mileage -- a very small mileage sample, reported as such. THE START-UP CLICK: A QUESTION THE DATA CANNOT SETTLE 358 complaints pair a click with a no-start or no-crank. Only 11 say the click was a SINGLE click and only 13 say it was rapid, repeated or machine-gun clicking. Of the 11 single-click reports, 5 name a battery and 4 name a starter or solenoid; of the 13 rapid reports, 8 name a battery and 2 name a starter or solenoid. THESE SAMPLES ARE FAR TOO SMALL TO SEPARATE THE TWO CAUSES, and in both of them the battery is named more often. The article states this plainly and sends readers to the load test instead of the click count. MANUFACTURER COMMUNICATIONS Full-text search of the TSB file for click | clicking | clicks returns 1,185 distinct NHTSA document ids and 1,228 distinct make + bulletin-number pairs. Largest by make: Ford 149, Chevrolet 121, GMC 86, Honda 83, Cadillac 78, Audi 54. -------------------------------------------------------------------------- LIMITATIONS -------------------------------------------------------------------------- 1. Complaints are self-reported, unverified, and not normalized by how many vehicles are on the road. Counts by make or model say nothing about reliability and no such claim is made in the article. 2. Group membership is a word match on the owner's narrative, not a technician's finding. Where a group is small (heat shield, 7; oil pressure, 53) the article says the sample is small rather than generalizing from it. 3. The heat-shield count being near zero is a reporting artifact as much as anything: owners file safety complaints about things they think are dangerous, and a rattling heat shield is not. It does not mean loose heat shields are rare on the road. 4. Crash, fire and tow are structured Y/N fields filled in at intake. A flag means both facts appear in one report, never that the click caused the crash. -------------------------------------------------------------------------- PRIMARY DOCUMENTS OPENED FOR THIS ARTICLE (18 September 2026) -------------------------------------------------------------------------- Hyundai TSB 17-ST-002, February 2017, MDPS coupling warranty extension TXX7 (NHTSA 10110922) https://static.nhtsa.gov/odi/tsbs/2017/MC-10110922-9999.pdf Kia TSB CHA044 Rev 2, May 2014, MDPS flexible coupling replacement (NHTSA 10109691) https://static.nhtsa.gov/odi/tsbs/2014/MC-10109691-9999.pdf Honda Tech Line Summary ATS30070, version 1 dated 19 December 2025, 2024-26 Prologue (NHTSA 11026973) https://static.nhtsa.gov/odi/tsbs/2026/MC-11026973-0001.pdf Ford SSM 49264, 4 November 2020, 2019-2020 Ranger blend door (NHTSA 10181907) https://static.nhtsa.gov/odi/tsbs/2020/MC-10181907-0001.pdf General Motors PIT6222, published 21 August 2024, rear HVAC mode door (NHTSA 11007762) https://static.nhtsa.gov/odi/tsbs/2024/MC-11007762-0001.pdf General Motors 17-NA-047, February 2017, wheel speed sensor encoder ring debris (NHTSA 10116633) https://static.nhtsa.gov/odi/tsbs/2017/MC-10116633-9999.pdf The recall and bulletin documents already cited in the article's sources list were carried over from the 13 September 2026 pass and not re-opened this week; the six above were opened as PDFs on 18 September 2026. ------------------------------------------------------------------------------ CASE SENSITIVITY — SECTION ADDED 19 SEPTEMBER 2026 ------------------------------------------------------------------------------ NHTSA stores complaint narratives in ALL CAPS for records filed through roughly 2020 and in mixed case for records filed from 2021 on. A case-SENSITIVE search of the narrative field loses about 92% of the complaints filed after 2020, so this is the single largest way an analysis of this database can go wrong while still looking finished. Every query behind this file is case-insensitive. SQLite's LIKE and its FTS5 full-text index both ignore case by default, and no query in this work uses GLOB or any case-sensitive comparison. This was not taken on trust. On 19 September 2026 every published folder was re-checked against the same database, comparing the figures here with what a case-sensitive query would have returned. The result: the complaint set published here (7,828) is LARGER than the absolute ceiling a case-sensitive search of the same terms could have returned (1,252 narratives). A case-sensitive query cannot produce a bigger set than itself, so the bug is ruled out for this folder by arithmetic, not by assurance.