# Long COVID, MCAS, Mold, Heavy Metals, SIBO and Mitochondrial Collapse: A Case That Shows What Happens When We Treat the Target but Forget the Host

> A complex case combining Long COVID, ME/CFS physiology, hyperadrenergic POTS, MCAS-type hyperreactivity, recurrent SIBO and methanogen overgrowth, mycotoxin findings, elevated toxic metals and profound mitochondrial respiratory-chain impairment — read as a single host-capacity failure rather than a list of separate targets.

- **Author:** Mohammed Attallah (BiomeLogic)
- **Published:** 2026-08-26
- **Category:** Case Reasoning
- **Tags:** long-covid, me-cfs, hyperadrenergic-pots, mcas, sibo, methanogens, hydrogen-sulfide, mold, mycotoxins, heavy-metals, mitochondrial-dysfunction, host-capacity-model, case-reasoning
- **Canonical URL:** https://biomelogic.net/articles/long-covid-mcas-mold-metals-mitochondrial-collapse-case
- **License:** CC BY-NC 4.0 — please cite "Mohammed Attallah, BiomeLogic" with link to https://biomelogic.net/articles/long-covid-mcas-mold-metals-mitochondrial-collapse-case.

---

I have been working through an extraordinarily complex case involving Long COVID, severe ME/CFS physiology, hyperadrenergic POTS, MCAS-type hyperreactivity, recurrent SIBO and intestinal methanogen overgrowth, hydrogen sulfide-producing bacteria, substantial intestinal inflammation, mold and mycotoxin findings, elevated toxic metals, impaired digestion, severe nutrient depletion, progressive muscle loss and profound mitochondrial respiratory-chain dysfunction. What makes this case important is not one abnormal test. It is what happens when the entire chronology is reconstructed and the findings are placed beside one another as parts of the same biological system.

If we look only at the breath test, we see methane and think about eliminating methanogens. If we look only at the microbiome, we see dysbiosis and think about antimicrobials. If we look only at mycotoxin testing, we think about antifungals or binders. If we look only at MCAS, we think about histamine. If we look only at POTS, we think about heart rate and blood pressure. If we look only at environmental toxicants, we think about detoxification. But the human body does not operate as a collection of separate laboratory compartments. Mitochondria, intestinal epithelial cells, autonomic nerves, mast cells, vascular endothelium, bile acids, pancreatic enzymes, microbial communities, liver metabolism, nutrient status and environmental exposures continuously influence one another. That interconnectedness is the central point of this case.

The mitochondrial findings immediately changed how I viewed everything else. MitoSwab testing showed Complex IV activity at only 22% of the normal mean, with an absolute activity of 0.069, while Complex II plus III activity was only 16% of the normal mean at 0.014. Yet citrate synthase was 68.13, approximately 563% of the normal mean, and the Complex I to Complex IV ratio reached 51.0 compared with a reference range of 12.1 to 30.3. Citrate synthase is commonly used as a proxy for mitochondrial content, so this is not simply a picture of a cell failing to make mitochondria. It suggests a system attempting to compensate by increasing mitochondrial mass while respiratory capacity remains severely constrained.

That distinction is fundamental. Complex IV, cytochrome c oxidase, sits at the end of the mitochondrial electron-transport chain. Electrons arriving from upstream respiratory complexes ultimately have to reach Complex IV, where oxygen serves as the terminal electron acceptor. The electron-transport chain creates an electrochemical proton gradient across the inner mitochondrial membrane, and ATP synthase uses that gradient to convert ADP and phosphate into ATP. If respiratory-chain throughput is severely restricted, simply producing more mitochondrial mass cannot completely overcome the bottleneck. The cell can therefore be working harder to compensate while still producing inadequate usable energy.

ATP is often discussed as if it merely determines whether someone feels energetic. It does much more than that. ATP maintains ion gradients, powers Na+/K+-ATPase, drives active membrane transport, supports protein synthesis, helps maintain membrane architecture, participates in hepatic and renal transport, supports tissue repair and supplies the energetic currency required for cells to preserve homeostasis. A person can therefore have an enormous problem with ATP availability even if routine blood chemistry does not look catastrophic. This case is a good example of why severe physiological dysfunction can coexist with relatively ordinary conventional blood testing early in the course. The reports specifically noted that routine hematology and chemistry can remain unrevealing while autonomic dysmotility, small-bowel gas production, altered microbial metabolism, mucosal immune activity and microvascular abnormalities are occurring in other compartments.

Then the gastrointestinal findings begin intersecting directly with the mitochondrial findings. The microbiome showed Desulfovibrio around 0.94% and Bilophila wadsworthia around 0.41%, with later analysis reporting Desulfovibrio around 1.73 × 10⁹ CFU/g. Both organisms belong to metabolic niches capable of contributing to hydrogen sulfide generation.

Hydrogen sulfide is not merely an intestinal gas. At physiological concentrations it participates in signaling, but excessive H2S can inhibit mitochondrial respiration. One of its important molecular targets is Complex IV itself, where H2S can interact with the heme a3-CuB catalytic center responsible for the final reduction of oxygen. Excessive inhibition at this point can reduce electron transfer, impair proton pumping, reduce mitochondrial membrane potential and constrain oxidative phosphorylation. Now place that next to the objective mitochondrial result: Complex IV activity at 22% of the normal mean. The article is not arguing that a stool organism alone proves the cause of the Complex IV abnormality. The important observation is that the microbial metabolic capability and the host respiratory abnormality converge on the same biological bottleneck.

But the deeper question is why this microbial environment developed in the first place. The case contains severe post-viral autonomic dysfunction. During NASA Lean testing, the supine heart rate was approximately 72 beats per minute; upright positioning produced profound intolerance and collapse, with heart rate subsequently reaching approximately 140 beats per minute. The larger history describes orthostatic intolerance, post-exertional malaise, food-triggered reactions and severe autonomic instability.

The small intestine depends heavily on coordinated neural and muscular activity. During fasting, the migrating motor complex generates cyclic activity, with Phase III producing strong propagating contractions that clear residual food, secretions and microorganisms. If post-viral autonomic or enteric signaling becomes impaired, small-intestinal clearance can slow. Residence time increases, bacteria and archaea have more opportunity to proliferate, fermentation increases, and the resulting microbial products can then influence epithelial, immune and neural function. The case analysis therefore identified post-viral neuroimmune/autonomic dysmotility as an upstream candidate driver rather than treating recurrent SIBO as an isolated infection.

That concept becomes especially important when the breath-test trajectory is examined. Earlier testing demonstrated hydrogen-predominant SIBO, with a hydrogen rise around 65 ppm and a peak around 126 ppm. Later, methane reached approximately 90.11 ppm. In other words, this was not a static microbiome problem; the fermentation phenotype itself changed over time.

Methane is also physiologically relevant because intestinal methane production has been associated with slower intestinal transit. That can create a self-reinforcing loop: impaired motility favors methanogenic expansion, methane is associated with further slowing of transit, slower transit increases microbial residence time, and the ecological niche becomes increasingly stable. If the upstream motility defect remains unresolved, repeated microbial eradication may reduce abundance temporarily without eliminating the physiology that keeps selecting for recurrence.

That appears particularly important here because the patient underwent repeated treatment. Rifaximin 550 mg three times daily for 14 days was used more than once, along with allicin. Later regimens included oregano, goldenseal, nystatin, itraconazole, monolaurin and multiple botanical antimicrobial preparations. The broader treatment history also included IV ozone, Ten-Pass ozone, EBOO, methylene blue, Myers infusions, high-dose vitamin C, phosphatidylcholine infusions, Cerebrolysin, Poly MVA, curcumin, resveratrol, BPC-157 and multiple peptides. The documented response to rifaximin was transient gastrointestinal relief followed by recurrence, which is exactly why the upstream ecology and host physiology became more important than simply repeating another kill strategy.

At the same time, the intestinal mucosa was becoming increasingly abnormal. Fecal calprotectin moved through approximately 25, 120, 80.9, 57, 202 and 122 mcg/g. Fecal lactoferrin became positive. Secretory IgA reached greater than 7,500 and later remained around 5,310 µg/g. Another report quantified fecal lactoferrin at 8.39 µg/mL, above the laboratory cutoff. These findings indicate a mucosal environment with significant immune activity rather than a simple overgrowth occurring against an otherwise healthy epithelial surface.

That matters because antimicrobial treatment is not biologically neutral. When Gram-negative organisms are damaged, the host may be exposed to additional lipopolysaccharide, lipoproteins, peptidoglycan fragments, nucleic acids and other microbial-associated molecules. LPS can interact with LPS-binding protein, CD14 and the TLR4-MD2 complex, activating MyD88- and TRIF-associated signaling. Downstream activation of IRAKs, TRAF6, TAK1 and the IKK complex can promote IκB degradation and permit NF-kappaB signaling to alter inflammatory transcription. The point is not that antimicrobial therapy should never be used. The point is that the biological consequences of microbial disruption can be very different in a resilient host compared with a host already experiencing mucosal inflammation, autonomic instability, mast-cell hyperreactivity and impaired mitochondrial respiration.

The systemic immune profile reinforces that concern. The Long Hauler Index was reported at 19.43 compared with a reference below 0.70. IFN-gamma was approximately 183.2 pg/mL, IL-6 32.8 pg/mL, IL-4 61.2 pg/mL and IL-13 53.3 pg/mL, with additional elevation of IL-8, IL-10 and TNF-alpha. Intermediate CD14+CD16+ monocytes were approximately 14.27%. Interestingly, the same reports describe total IgE around 24 kU/L with food, mold and alpha-gal panels negative. That is relevant because it argues against interpreting every reaction as conventional allergen-specific IgE sensitization and supports considering broader inflammatory, autonomic and non-IgE mechanisms in the phenotype.

The microbial ecosystem itself also shows why simply labeling organisms “good” or “bad” misses what is happening. Bacteroidetes reached approximately 73%, with Bacteroides around 64.67%. Firmicutes were around 18%. Bifidobacterium fell to approximately 0.02%, Roseburia to 0.07%, Eubacterium to 0.09% and Coprococcus to 0.01%. Faecalibacterium remained around 10.42%, which superficially looks reassuring because it is a major butyrate producer, but the deeper issue was loss of functional redundancy.

A healthy ecosystem does not depend on one organism to carry an entire metabolic function. Bifidobacteria generate acetate and lactate that can be used by other organisms. Faecalibacterium, Roseburia, Eubacterium and Coprococcus participate in overlapping butyrate-producing networks. If Bifidobacterium falls to 0.02% and multiple butyrate-producing partners collapse while Faecalibacterium remains relatively abundant, relative abundance alone does not tell us actual metabolic flux. Cross-feeding substrate may be missing. The reports explicitly describe this as a collapse of functional redundancy rather than a total absence of butyrate-producing organisms.

That matters for the host because butyrate is more than a microbial metabolite. Colonocytes can oxidize butyrate as a major energy substrate. Normal colonocyte oxidative metabolism consumes oxygen and helps preserve the low-oxygen environment near the luminal surface that favors obligate anaerobes. If butyrate-producing networks and epithelial oxidative metabolism become unstable, the ecological consequences can extend beyond one metabolite. Oxygen availability, epithelial metabolism and microbial selection can all begin shifting together.

The diet then became another major selection pressure. At one stage, food intake had deteriorated to only two or three puréed vegetables in tiny amounts along with defatted chicken broth. Under extreme restriction of microbiota-accessible carbohydrates, organisms that depend on diverse fermentable substrate lose their ecological support. Bacteroides, however, possess extensive polysaccharide-utilization loci and can access a broad range of glycans. Under substrate scarcity, some species can increasingly use host-derived mucosal glycans containing compounds such as N-acetylglucosamine, fucose and sialic acid. The concern is that a microbiome starved of dietary substrate can increasingly turn toward the mucus environment itself.

That creates a potentially destructive ecological sequence: severe dietary restriction reduces substrate diversity, cross-feeding networks collapse, microbial redundancy decreases, mucin-utilizing organisms gain a survival advantage, the mucus barrier becomes more vulnerable, microbial antigens move closer to the epithelial surface and mucosal immune activation intensifies. If additional antimicrobial pressure is applied during that process, the host may lose even more ecological resilience.

Digestive physiology provides another part of the explanation. Pancreatic elastase, which had previously been greater than 500 µg/g, later fell to approximately 160 µg/g before subsequently returning to 677 µg/g. That fluctuation is important because it suggests that digestive capacity itself was dynamic rather than permanently fixed. Stool β-glucuronidase also reached approximately 3,228 U/mL, above the reported reference of 2,486, while conjugated bile acids were markedly elevated: TCDCA approximately 1,040 ng/g, TCA around 1,020 ng/g and GCA around 696 ng/g.

Bile acids are not simply detergents for absorbing fat. They are endocrine-like signaling molecules interacting with receptors such as FXR and TGR5 and influencing hepatic bile synthesis, motility, glucose metabolism, immune function and microbial ecology. Excess conjugated bile acids arriving distally can therefore represent both a digestive abnormality and an ecological signal. Bilophila is particularly relevant because it is bile tolerant and can use sulfur-containing substrates derived from taurine-conjugated bile acids. In a system already enriched for H2S-associated organisms, bile handling becomes part of the microbial ecology rather than a separate digestive footnote.

β-Glucuronidase adds another dimension. Hepatic glucuronidation attaches glucuronic acid to many endogenous compounds and xenobiotics to facilitate elimination. Microbial β-glucuronidase can remove that glucuronide group in the intestine, potentially allowing some compounds to be reabsorbed and re-enter enterohepatic circulation. That creates another example of why the liver and microbiome cannot always be analyzed independently.

The vascular findings add still another energetic constraint. The reports contain Stage 3.5 out of 4 amyloid fibrin microclot findings and endothelial casts, along with fibrinogen around 436 mg/dL, Protein S activity around 54%, PAI-1 4G/4G, anticardiolipin IgM rising from approximately 16 to 27 MPL U/mL and phosphatidylserine/prothrombin antibodies. At the same time, the broader coagulation workup included normal D-dimer/FDP and negative lupus-anticoagulant testing, meaning this pattern should not simply be collapsed into a conventional diagnosis of antiphospholipid syndrome.

The mechanistic question raised by the microvascular findings is tissue delivery. Peripheral pulse oximetry can read 98% or 99% while oxygen delivery at the microcirculatory level is still imperfect. Mitochondria do not merely need oxygen to be present in arterial blood; oxygen and metabolic substrate have to reach the tissue. If microvascular delivery is compromised while respiratory-chain activity is already abnormal, the host is being squeezed from both directions: less efficient delivery upstream and impaired oxidative utilization downstream. The report explicitly connected the Stage 3.5 microclot pattern with the possibility of functional tissue hypoxia despite normal peripheral oxygen saturation.

Then I looked at the environmental toxicant profile. Urinary testing showed barium at 7.43 µg/g compared with a laboratory reference of 5.59 or lower, palladium at 1.42 µg/g compared with 0.20 or lower, thallium at 1.13 µg/g compared with 0.43 or lower and uranium at 1.24 µg/g compared with 0.04 or lower. Within this testing framework, the uranium result was approximately 31 times the laboratory upper reference limit.

These compounds should not simply be grouped together under the word “toxins,” because their chemistry is different. Barium has a particularly important relationship with potassium-channel physiology. Ba²⁺ can enter potassium-channel pores and interfere with K+ conductance. Potassium currents are essential for membrane repolarization, and membrane repolarization is required for repetitive signaling in neurons and muscle. Intestinal smooth muscle depends on coordinated calcium and potassium conductances to move between contraction and relaxation. Disturbances in potassium-channel activity can therefore affect electrical excitability and motility.

Thallium behaves differently. Tl+ has chemical similarities to K+ that allow it to exploit some potassium-handling pathways. Once inside cells, thallium can interact with sulfhydryl-containing proteins, alter ion homeostasis and interfere with mitochondrial processes. In a host already demonstrating profound respiratory-chain impairment, another stressor converging on mitochondrial and membrane physiology becomes mechanistically relevant.

Uranium introduces yet another chemistry. In its uranyl form, UO2²⁺ has a strong affinity for oxygen-containing ligands, including phosphate and carbonate. Phosphate-rich structures include the head groups of membrane phospholipids. Cell membranes are not simply structural walls; they are platforms containing receptors, ion channels, ATPases, transport proteins and signaling complexes. Interference with membrane organization therefore has the potential to influence much more than membrane integrity alone.

This brings us to one of the most important concepts in the entire case: detoxification itself requires physiological capacity. The body has to absorb, bind, transform, compartmentalize, transport and eliminate compounds. The liver has to process substances and move them into bile. The kidneys have to filter and perform energy-dependent tubular transport. Cells have to operate membrane exporters. P-glycoprotein, MRP2, MRP4 and BCRP belong to ATP-binding cassette transporter systems whose transport cycles require ATP.

That means the same host being asked to clear environmental compounds has Complex IV at 22% and Complex II plus III at 16%. ATP is required to operate the very transport machinery being asked to perform clearance. This does not mean ATP is the only determinant of toxicant elimination, but it means bioenergetics cannot be separated from the discussion. If oxidative ATP generation is severely constrained, aggressive mobilization or treatment may increase demand before transport, repair and elimination capacity have been adequately considered.

The mycotoxin findings add a different set of molecular targets. Fumonisin B1 was approximately 10.35 ng/g compared with the laboratory reference of 6.13 or lower, Satratoxin H approximately 0.62 ng/g compared with 0.18 or lower, and Zearalenone approximately 1.79 ng/g compared with 0.67 or lower. Aflatoxin B1 was approximately 3.93 ng/g.

Fumonisin B1 is mechanistically interesting because it inhibits ceramide synthase. Ceramide synthases participate in sphingolipid metabolism, and sphingolipids serve both structural and signaling roles in cellular membranes. They influence membrane organization, epithelial integrity, apoptosis, inflammatory signaling and cellular stress responses. In an intestinal mucosa already demonstrating inflammatory injury, interference with sphingolipid metabolism becomes particularly relevant.

Macrocyclic trichothecenes such as Satratoxin H interact with the eukaryotic ribosomal machinery and can trigger the ribotoxic stress response. Ribosomal function matters enormously in the intestinal epithelium because epithelial cells require continuous protein synthesis to replace transporters, enzymes, receptors, cytoskeletal proteins, mucins and tight-junction components. A mucosa with calprotectin reaching 202 mcg/g and positive lactoferrin is already under repair pressure; impaired translational capacity represents another potential burden on that repair process.

Zearalenone acts through yet another pathway. Its structure resembles 17-beta estradiol sufficiently to interact with estrogen receptors ER-alpha and ER-beta. These receptors are transcriptional regulators; ligand binding alters receptor conformation and changes gene transcription through estrogen-response elements and interacting co-regulatory pathways. That means zearalenone is not simply a generic “mold toxin.” Its biology intersects with endocrine signaling, epithelial gene expression and immune regulation.

This is precisely why mechanistic analysis matters. Barium influences potassium-channel physiology. Thallium behaves in part as a potassium mimic and can stress mitochondrial systems. Uranyl chemistry targets phosphate-rich structures. Fumonisin interferes with sphingolipid synthesis. Trichothecenes target ribosomal function. Zearalenone interacts with estrogen receptors. Different molecules have different primary targets, but the downstream consequences can converge on a much smaller set of host vulnerabilities: membrane integrity, mitochondrial energetics, ion gradients, protein synthesis, epithelial repair, immune activation and intestinal motility.

Meanwhile the host was progressively losing the raw materials required to compensate. Vitamin D moved from approximately 35 to 24 to 12 to 13.6 and ultimately 7.7 ng/mL. Intracellular micronutrient testing placed vitamin B3 or niacin around the deficiency threshold at 80%, while chromium was 41%, copper 45%, glutamine 42%, serine 32% and vitamin K2 36% in the borderline range reported by that assay.

Niacin is important because it supplies precursors for NAD+, one of the central redox cofactors in cellular metabolism. Glycolysis requires NAD+. Pyruvate dehydrogenase generates NADH from NAD+. Multiple TCA-cycle enzymes depend on NAD+/NADH cycling. Fatty-acid oxidation relies on the same redox system, and NAD+ is also consumed by enzymes such as sirtuins and PARPs. In other words, a host already struggling with respiratory-chain output was also showing weakness in one of the nutrient systems required to feed and regulate energy metabolism.

The severe vitamin D decline occurred alongside floating yellow/orange stools, altered bile physiology, severe dietary fat restriction and coagulation abnormalities. PT reached approximately 16.2 seconds, aPTT 35.4 seconds and INR 1.3 while thrombin time remained around 20.3 seconds. The analysis therefore raised impaired fat-soluble vitamin handling, including possible vitamin K insufficiency, as a mechanistic hypothesis rather than assuming that every coagulation abnormality represented a primary clotting disorder.

The loss of body reserve is equally important. Serum creatinine declined into approximately the 0.52 to 0.58 mg/dL range, CK had been below approximately 25 U/L, urinary 3-methylhistidine reached roughly 575 µmol/g creatinine and body weight fell from approximately 130 pounds to around 110 to 112 pounds. This was occurring in the context of an extremely restricted diet.

Skeletal muscle is not merely tissue used for movement. It is a major metabolic reservoir of amino acids. During prolonged nutritional and energetic stress, the body can increase muscle proteolysis to release amino acids for gluconeogenesis, acute-phase proteins, enzymes and other essential processes. Rising 3-methylhistidine, falling body weight and low creatinine therefore become much more meaningful when interpreted together than when any one marker is viewed in isolation.

Even the liver began showing a changing trajectory. Baseline ALT was approximately 26 U/L and later increased to 53 and then 61 U/L during a period involving oral itraconazole, concentrated botanical antimicrobial preparations and a substantial total intervention burden. Bilirubin and alkaline phosphatase remained normal, making the pattern more hepatocellular than cholestatic in the analysis.

That treatment chronology is one of the most important lessons from this case. A medication or antimicrobial can have a perfectly rational molecular target and still be poorly sequenced for the physiological state of the person receiving it. The question “Can this compound suppress this organism?” is different from the question “Does this host currently have the mitochondrial, hepatic, nutritional, mucosal, autonomic and vascular reserve to tolerate what happens when I use it?”

That is the concept I mean by Host Capacity.

Host Capacity does not mean ignoring SIBO, mold, mycotoxins, methane, hydrogen sulfide, environmental metals or inflammation. It means refusing to interpret them outside the physiological condition of the person carrying them. Methane around 90 ppm matters, but so does why intestinal transit failed. Desulfovibrio matters, but so does what selected for it and whether its metabolites converge on an existing mitochondrial bottleneck. Mycotoxins matter, but so do ribosomal function, sphingolipid biology, membrane integrity and the host’s ability to eliminate compounds. Heavy metals matter, but so do ATP-dependent transport, renal handling and mitochondrial reserve. Dysbiosis matters, but so do substrate starvation, bile acids, mucus utilization, cross-feeding and epithelial metabolism.

When all the data are viewed together, the picture becomes difficult to reduce to one diagnosis: Complex IV at 22%, Complex II plus III at 16%, citrate synthase at 563%, a Long Hauler Index of 19.43, IFN-gamma around 183.2 pg/mL, methane around 90.11 ppm, calprotectin reaching 202 mcg/g, positive fecal lactoferrin, severe depletion of Bifidobacterium and multiple butyrate-producing partners, abnormal conjugated bile acids, transient pancreatic elastase around 160 µg/g, vitamin D eventually reaching 7.7 ng/mL, substantial weight and muscle loss, ALT rising from 26 to 61 U/L, microvascular abnormalities, and environmental toxicant and mycotoxin findings layered on top of all of it.

In a host like this, I do not think the correct first question is automatically, “What else can we kill?” Nor is it automatically, “What else can we detox?” The more fundamental question is which physiological bottlenecks are currently preventing the system from recovering, which interventions have already increased the burden, what capacities remain intact, and what must be rebuilt so that future treatment becomes tolerable rather than simply adding another stressor.

That is also why I am cautious about automatically labeling every major deterioration after treatment as “die-off” or a “Herxheimer reaction.” Sometimes microbial killing may contribute to transient inflammatory symptoms. But sometimes worsening may represent treatment intolerance, ecological disruption, hepatic stress, inadequate nutrition, autonomic destabilization or simply a host being asked to perform more biological work than its current reserve can support.

For me, that is the central lesson of this entire case. Complex chronic illness requires more than identifying abnormal targets. It requires reconstructing the sequence connecting infection, autonomic function, motility, microbial ecology, mucosal immunity, mitochondrial energy production, vascular delivery, digestion, environmental exposure, nutrient reserve and treatment response.

The question is not simply whether we can treat the target. The question is whether we understand why the target emerged, what it is doing to the rest of the system, and whether the host currently has the capacity to tolerate what we are about to do.

---

*Source of record: https://biomelogic.net/articles/long-covid-mcas-mold-metals-mitochondrial-collapse-case. This plain-text endpoint is provided for AI retrieval and citation systems (ChatGPT, Perplexity, Claude, Gemini, Bing Copilot).*